update 127的pol_cell算白点

This reverts commit 01247bccb9.

AIEngineModule/include_base/AI_Factory.h

@@ -174,7 +174,8 @@ public:
 
     std::shared_ptr<AIModel_Base> AI_defect_zf; // L127 L255
 
-    std::shared_ptr<AIModel_Base> AI_defect_bq; // L127 L255
+    std::shared_ptr<AIModel_Base> AI_defect_bq;
+    std::shared_ptr<AIModel_Base> AI_defect_Tag;
 
     std::shared_ptr<AIModel_Base> AI_defect_127Cell; // L127 L255
 
@@ -191,6 +192,8 @@ public:
     std::shared_ptr<AIModel_Base> AI_defect_MarkLine; // 缺pol检测
     std::shared_ptr<AIModel_Base> AI_defect_Edge_QX;  // 边缘缺陷检测
 
+     std::shared_ptr<AIModel_Base> AI_defect_RLine_512;  // 弱线检测模型 512*512
+
 private:
     // 模型是否都加载完成了。
     bool m_bInitSucc = false;

AIEngineModule/src/AIModel_Impl.cpp

@@ -265,14 +265,15 @@ int AIModel_Impl::AIDet(const cv::Mat &inImg, cv::Mat &outimg)
     {
         outimg.create(m_pNode_output_0->height, m_pNode_output_0->width, outType);
     }
+    // 推理前清零输出：任何一步失败时，保证输出全黑而非未初始化垃圾数据
+    outimg.setTo(0);
     uchar *p_outdata_1 = (uchar *)outimg.data;
     if (!m_bInitSuccess)
     {
         return 2;
     }
 
-    AI_Det_In_1_Out_1(m_pNode_input_0, m_pNode_output_0, p_indata_0, p_outdata_1);
-    return 0;
+    return AI_Det_In_1_Out_1(m_pNode_input_0, m_pNode_output_0, p_indata_0, p_outdata_1);
 }
 
 int AIModel_Impl::AIDet(const cv::Mat &inImg, cv::Mat &outimg0, cv::Mat &outimg1)
@@ -553,38 +554,89 @@ int AIModel_Impl::GetEngineInfo1(std::shared_ptr<GPU_Engine> &pgpuengine)
 
 int AIModel_Impl::AI_Det_In_1_Out_1(Node_Config *pConfig_in, Node_Config *pConfig_out, const unsigned char *p_indata_0, unsigned char *p_outdata_1)
 {
-    // printf("=== s1 ");
     std::shared_ptr<Det_GPU_Stram> pdetStream;
     if (GetStream(pdetStream) != 0 || !pdetStream)
     {
         return -1;
     }
-    //  printf("=== s2 ");
     std::lock_guard<std::mutex> lock(pdetStream->AI_mutex);
-    // printf("  ss g %d s %d --  ", pdetStream->nGPUIdx, pdetStream->cuda_stream->nstreamIdx);
-    // // 设置 显卡ID
+    // 设置 显卡ID
     if (!EnsureCudaDevice(pdetStream->nGPUIdx))
     {
         return -1;
     }
 
-    cudaMemcpyAsync(pdetStream->cuda_stream->pstreamNode_input_0->gpu_ImgData,
-                    p_indata_0, pConfig_in->ucharsize, cudaMemcpyHostToDevice, pdetStream->cuda_stream->stream);
-    // // //  在显存中 图片数据从 uchar 转到 float
-    Cuda_ucharToFloat_stream((unsigned char *)pdetStream->cuda_stream->pstreamNode_input_0->gpu_ImgData,
-                             (float *)pdetStream->cuda_stream->pstreamNode_input_0->gpu_buffers, pConfig_in->datalength, pdetStream->cuda_stream->stream);
+    cudaStream_t stream = pdetStream->cuda_stream->stream;
+    void *gpu_in_ImgData = pdetStream->cuda_stream->pstreamNode_input_0->gpu_ImgData;
+    void *gpu_in_buffers = pdetStream->cuda_stream->pstreamNode_input_0->gpu_buffers;
+    void *gpu_out_buffers = pdetStream->cuda_stream->pstreamNode_output_0->gpu_buffers;
+    void *gpu_out_ImgData = pdetStream->cuda_stream->pstreamNode_output_0->gpu_ImgData;
 
-    pdetStream->cuda_stream->context->setTensorAddress(pConfig_in->name.c_str(), pdetStream->cuda_stream->pstreamNode_input_0->gpu_buffers);
-    pdetStream->cuda_stream->context->setTensorAddress(pConfig_out->name.c_str(), pdetStream->cuda_stream->pstreamNode_output_0->gpu_buffers);
+    cudaError_t cerr;
 
-    pdetStream->cuda_stream->context->enqueueV3(pdetStream->cuda_stream->stream);
+    // 推理前先把输出显存清零：任何一步失败时，保证输出全黑而非未初始化垃圾数据
+    cerr = cudaMemsetAsync(gpu_out_ImgData, 0, pConfig_out->ucharsize, stream);
+    if (cerr != cudaSuccess)
+    {
+        std::fprintf(stderr, "[AIEngine] cudaMemsetAsync(out) failed: %s\n", cudaGetErrorString(cerr));
+        return -1;
+    }
 
-    Cuda_FloatTouchar_stream((float *)pdetStream->cuda_stream->pstreamNode_output_0->gpu_buffers,
-                             (unsigned char *)pdetStream->cuda_stream->pstreamNode_output_0->gpu_ImgData, pConfig_out->datalength, pdetStream->cuda_stream->stream);
-    cudaMemcpyAsync(p_outdata_1, pdetStream->cuda_stream->pstreamNode_output_0->gpu_ImgData,
-                    pConfig_out->ucharsize, cudaMemcpyDeviceToHost, pdetStream->cuda_stream->stream);
-    cudaStreamSynchronize(pdetStream->cuda_stream->stream);
-    // printf("**ee g %d s %d **  \n", pdetStream->nGPUIdx, pdetStream->cuda_stream->nstreamIdx);
+    // 输入数据拷贝到显存
+    cerr = cudaMemcpyAsync(gpu_in_ImgData, p_indata_0, pConfig_in->ucharsize, cudaMemcpyHostToDevice, stream);
+    if (cerr != cudaSuccess)
+    {
+        std::fprintf(stderr, "[AIEngine] cudaMemcpyAsync(in) failed: %s\n", cudaGetErrorString(cerr));
+        return -1;
+    }
+
+    // 在显存中 图片数据从 uchar 转到 float
+    Cuda_ucharToFloat_stream((unsigned char *)gpu_in_ImgData,
+                             (float *)gpu_in_buffers, pConfig_in->datalength, stream);
+
+    if (!pdetStream->cuda_stream->context->setTensorAddress(pConfig_in->name.c_str(), gpu_in_buffers))
+    {
+        std::fprintf(stderr, "[AIEngine] setTensorAddress(in) failed\n");
+        return -1;
+    }
+    if (!pdetStream->cuda_stream->context->setTensorAddress(pConfig_out->name.c_str(), gpu_out_buffers))
+    {
+        std::fprintf(stderr, "[AIEngine] setTensorAddress(out) failed\n");
+        return -1;
+    }
+
+    if (!pdetStream->cuda_stream->context->enqueueV3(stream))
+    {
+        std::fprintf(stderr, "[AIEngine] enqueueV3 failed\n");
+        return -1;
+    }
+
+    // float 转到 uchar
+    Cuda_FloatTouchar_stream((float *)gpu_out_buffers,
+                             (unsigned char *)gpu_out_ImgData, pConfig_out->datalength, stream);
+
+    cerr = cudaMemcpyAsync(p_outdata_1, gpu_out_ImgData,
+                           pConfig_out->ucharsize, cudaMemcpyDeviceToHost, stream);
+    if (cerr != cudaSuccess)
+    {
+        std::fprintf(stderr, "[AIEngine] cudaMemcpyAsync(out) failed: %s\n", cudaGetErrorString(cerr));
+        return -1;
+    }
+
+    cerr = cudaStreamSynchronize(stream);
+    if (cerr != cudaSuccess)
+    {
+        std::fprintf(stderr, "[AIEngine] cudaStreamSynchronize failed: %s\n", cudaGetErrorString(cerr));
+        return -1;
+    }
+
+    // kernel 启动失败等异步错误必须通过 cudaGetLastError 捕获
+    cerr = cudaGetLastError();
+    if (cerr != cudaSuccess)
+    {
+        std::fprintf(stderr, "[AIEngine] async cuda error: %s\n", cudaGetErrorString(cerr));
+        return -1;
+    }
     return 0;
 }
 

AIEngineModule/src/AI_Factory.cpp

@@ -143,7 +143,7 @@ int AIFactory::InitALLAIModle(GPU_Config gupconfig)
         boe_config.inputType = AIModel_Base::Input_CHW;
         AI_defect_zf->Init(boe_config);
     }
-    // 标签 检测
+    // 标签 定位
     if (!AI_defect_bq)
     {
         AI_defect_bq = AIModel_Base::GetInstance();
@@ -154,6 +154,17 @@ int AIFactory::InitALLAIModle(GPU_Config gupconfig)
         boe_config.inputType = AIModel_Base::Input_CHW;
         AI_defect_bq->Init(boe_config);
     }
+    // 标签 检测
+    if (!AI_defect_Tag)
+    {
+        AI_defect_Tag = AIModel_Base::GetInstance();
+        AIModel_Base::AIModelRun_Config boe_config;
+        boe_config.gpuconfig.copy(gupconfig);
+        boe_config.strPath = "/home/aidlux/BOE/UseModel_FOG/defect_Tag.engine";
+        boe_config.strName = "tag";
+        boe_config.inputType = AIModel_Base::Input_CHW;
+        AI_defect_Tag->Init(boe_config);
+    }
     // 字符 127cell
     if (!AI_defect_127Cell)
     {
@@ -253,6 +264,17 @@ int AIFactory::InitALLAIModle(GPU_Config gupconfig)
         boe_config.inputType = AIModel_Base::Input_CHW;
         boe_config.strName = "Edge_QX";
         AI_defect_Edge_QX->Init(boe_config);
+    }
+        // 弱线检测模型 512*512
+    if (!AI_defect_RLine_512)
+    {
+        AI_defect_RLine_512 = AIModel_Base::GetInstance();
+        AIModel_Base::AIModelRun_Config boe_config;
+        boe_config.gpuconfig.copy(gupconfig);
+        boe_config.strPath = "/home/aidlux/BOE/UseModel_FOG/FOG_L127line.engine";
+        boe_config.strName = "RLine_512";
+        boe_config.inputType = AIModel_Base::Input_CHW;
+        AI_defect_RLine_512->Init(boe_config);
     }
     m_bInitSucc = true;
     return 0;

AlgorithmModule/include/AI_Edge_QX_Det.h

@@ -45,7 +45,13 @@ public:
 	AI_Edge_QX_Det(/* args */);
 	~AI_Edge_QX_Det();
 
-	int Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDetConfig,std::shared_ptr<DetLog> pdetlog);
+	/// @brief 边缘缺陷检测，检测结果根据配置参数合并到 mask
+	/// @param img 检测原图
+	/// @param detmask 合并目标 mask（主检测）
+	/// @param pDetConfig 检测参数
+	/// @param pdetlog 日志
+	/// @param pDetmask2 可选：第二个合并目标 mask（如弱线检测mask），按与主检测相同规则合并
+	int Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDetConfig, std::shared_ptr<DetLog> pdetlog, cv::Mat *pDetmask2 = nullptr);
 
 private:
 	int SaveProcessImg(const cv::Mat &inImg, const cv::Mat &outImg, bool isALLzeros, DetConfig *pDetConfig, int idx = 0, int nruntype = 0);

AlgorithmModule/include/BlobBase.h

@@ -10,8 +10,8 @@ enum ERR_DOT_TYPE_ENUM
 #define Value_ERR_TYPE_2 128
 
 //-----------------------sxg added
-#define _MAX_ERROR_SCAN_LINE_PER_ROW 300
-#define _MAX_ERROR_DOT_BLOB 80
+#define _MAX_ERROR_SCAN_LINE_PER_ROW 1000
+#define _MAX_ERROR_DOT_BLOB 1000
 #define _MAX_MACRO_COUNT 4
 
 #define _MAX_ERR_CLASS ERR_TYPE_COUNT

AlgorithmModule/include/CheckErrorCodeDefine.hpp

@@ -26,7 +26,7 @@
 // resize  图片的 宽度
 #define RESIZE_IMAGE_WIDTH 1680
 
-// #define AI_Time
+// 结果图保存/绘制开关：已由 SAVE_RESIMG 宏改为 base 层配置项“是否存图绘制”（baseCheckData 中 itemCode=bDrawAndSave 的 isOpen），运行时生效，无需再改此处宏
 
 enum TEM_IMG_IDX_
 {

AlgorithmModule/include/CheckUtil.hpp

@@ -44,6 +44,7 @@ public:
 	static int SizeRect(cv::Rect &roi, int img_w, int img_h, int addw, int addh);
 	static int CalHj(const cv::Mat &img, const cv::Mat &mask, int b_value);
 	static int CalHj(const cv::Mat &img, const cv::Mat &mask, const cv::Mat &backgroundimg);
+	static int CalHjWeighted(const cv::Mat &img, const cv::Mat &mask, int b_value, float power = 2.0f);
 	static float CalRoi2RoiPre(cv::Rect rect1, cv::Rect rect2);
 	// 计算平均灰度
 	static float CalImgBrightness(cv::Mat imgRoi);

AlgorithmModule/include/ImageAllResult.h

@@ -46,8 +46,8 @@ public:
     struct Config_Update
     {
         bool bUpdate = false;
-        bool wait_DP= false; 
-        bool wait_UP= false; 
+        bool wait_DP = false;
+        bool wait_UP = false;
     };
 
     // AI MaskImg 的相关使用的参数信息。
@@ -173,6 +173,8 @@ public:
 
     std::shared_ptr<Image_AI_Det_Result> qx_DetAIResult;      // BLob 检测要用到的一些 参数 。
     std::shared_ptr<Image_AI_Det_Result> cell127_DetAIResult; // 127 cell 检测要用到的一些 参数 。
+    std::shared_ptr<Image_AI_Det_Result> bq_DetAIResult;      // BQ标签检测要用到的一些 参数 。
+    std::shared_ptr<Image_AI_Det_Result> RLine_512_DetAIResult; // 弱线512模型 检测要用到的一些 参数 。
 
     // 结果图片
     cv::Mat resultImg;
@@ -204,7 +206,6 @@ public:
 
     Config_Update config_update;
 
-
 public:
     DetStep m_step = DetStep_NotDet;
     std::string strBaseInfo = "";

AlgorithmModule/include/ImageDetConfig.h

@@ -40,6 +40,7 @@ struct QX_ERROR_INFO_
 	float grayDis;
 	float fUpIou;
 	float density;						 // Blob- 密度
+	bool bIsStandardLD;                  // 是否通过LD标准判定（跳过UP/DP的IOU检查）
 	std::vector<std::string> detLogList; // 检测日志
 	std::shared_ptr<DetLog> detlog;
 	std::vector<int> detRegionidxList; // 对应检测区域的 idx.
@@ -344,7 +345,11 @@ struct BQ_Result
 {
 	int nresult = 1;
 	bool bShield_BQ = true;			   // 是否屏蔽标签
-	std::vector<cv::Rect> pBQ_roiList;	   // 标签的区域
+	bool bBQ_AI_Det = false;		   // 是否启用标签独立AI检测
+	bool bsaveAll = false;			   // 是否保存所有的检测结果
+	std::vector<cv::Rect> pBQ_roiList;	   // 标签的原始区域
+	std::vector<cv::Rect> pBQ_expandedRoiList; // 标签的扩展后区域（用于AI检测）
+	std::vector<cv::Mat> pBQ_cropImages;	   // 标签扩展区域图像
 	std::vector<cv::Point> BQ_centerPoint; // 标签的中心点位置
 };
 struct CameraBaseResult

AlgorithmModule/include/ImageResultJudge.h

@@ -107,6 +107,15 @@ public:
     int UpdateConfig(std::string strcameraName);
     int DrawResult(std::shared_ptr<ImageAllResult> pImageResult);
 
+    // 盲失联合判断：主通道(ms_main_channel)检出的盲失(Other)若在过滤通道(ms_filter_channels)同位置存在缺陷
+    // (不限制缺陷类型)，则不算盲失(撤销NG)。同位置判断参考暗点联合分析：双方按 ms_expand 扩张后计算 IOU, 大于 ms_iou 认为同位置
+    int FilterMsBlindNG(std::shared_ptr<CameraResult> pCheck_Result);
+
+    // cell黑白点联合判断：主通道(cellbw_main_channel)检出的cell黑白点(Cell127 也算白点)在过滤通道
+    // (cellbw_filter_channels)同位置存在缺陷(不限制缺陷类型)时不算cell黑白点(撤销NG)
+    // 同位置判断参考暗点联合分析：双方按 cellbw_expand 扩张后计算 IOU, 大于 cellbw_iou 认为同位置
+    int FilterCellBwNG(std::shared_ptr<CameraResult> pCheck_Result);
+
 private:
     // 获取 缺陷在参数列表的位置
     int GetParamidx();
@@ -120,6 +129,9 @@ private:
     // 获取缺陷 对应的AI输入 输出图片，
     int GetAIDetImg(std::shared_ptr<ImageAllResult> pImageResult, cv::Point pcenter, cv::Mat &AI_InImg, cv::Mat &AI_OutImg);
 
+    // 绘制并保存结果图开关（base 配置 bDrawAndSave，替代 SAVE_RESIMG 宏）
+    bool NeedDrawAndSave();
+
     int sendTask(std::shared_ptr<DrawInfo> task);
 
     std::shared_ptr<ImageResultJudge::DrawInfo> GetTask();
@@ -132,6 +144,39 @@ private:
 
     int AnalysisResult_Pre(QX_ERROR_INFO_ *QX_info); // 弱化处理
 
+    // 撤销单个缺陷的NG结果：恢复为OK，移除对应NG小图并回退计数
+    int UndoNGResult(std::shared_ptr<ImageAllResult> pImageResult, int qxidx, QX_ERROR_INFO_ *QX_info);
+
+    // 跨通道同位置判断：当前缺陷(产品坐标系ROI)在过滤通道是否存在同位置缺陷(双方按 expand 扩张后计算IOU)
+    bool IsOverlapQxInFilterChannels(const std::vector<std::shared_ptr<ImageAllResult>> &imgResults,
+                                     const std::string &cur_channel_name,
+                                     const std::vector<std::string> &filter_channels,
+                                     const std::vector<int> &match_qx_types,
+                                     const cv::Rect &qx_product_roi,
+                                     int roi_expand, float iou_threshold,
+                                     const std::string &strTag,
+                                     std::shared_ptr<DetLog> plog);
+
+    // 判断盲失(产品坐标系ROI)在配置的过滤通道是否存在同位置缺陷(不限制类型，双方按 expand 扩张后计算IOU)
+    bool IsOverlapQxInMsFilterChannel(const std::vector<std::shared_ptr<ImageAllResult>> &imgResults,
+                                      const std::string &cur_channel_name,
+                                      const cv::Rect &qx_product_roi,
+                                      int roi_expand, float iou_threshold,
+                                      std::shared_ptr<DetLog> plog);
+
+    // 跨通道联合判断(通用)：主通道中指定类型的NG缺陷，若在过滤通道同位置存在指定类型的缺陷，则撤销该NG
+    int FilterCrossChannelNG(std::shared_ptr<CameraResult> pCheck_Result,
+                             const std::vector<std::string> &main_channels,
+                             const std::vector<std::string> &filter_channels,
+                             const std::vector<int> &main_qx_types,
+                             const std::vector<int> &match_qx_types,
+                             int roi_expand, float iou_threshold,
+                             const std::string &strTag,
+                             const std::string &strReason);
+
+    // 关联过滤源过滤：撤销与关联通道过滤源缺陷重叠的暗点NG结果(含单图判定的单点NG)
+    int FilterCorrelatedDirtyNG(std::shared_ptr<ImageAllResult> pImageResult);
+
     ChannelCheckFunction *GetChannelFuntion(std::string strChannelName); // 获得 通道的检测功能
 
     int ConfigTypeToQXAnalysis(int nconfigType);

AlgorithmModule/include/ImgCheckAnalysisy.hpp

@@ -73,6 +73,9 @@ public:
     // 获取检测库 状态信息   返回：CHECK_THREAD_RUN_STATUS
     int GetStatus();
 
+    // 计算blob的密度
+    int CalBlobDensity_QX();
+
     // 更新参数 pconfig 参数指针，nConfigType 需要更新的参数类型  返回：0 成功 其他异常
     int UpdateConfig(void *pconfig, int nConfigType);
 
@@ -190,6 +193,15 @@ private:
     int GetBlob_QX();
     // 获得 127cell blob
     int GetBlob_127cell();
+    // 获得 RLine_512 blob
+    int GetBlob_RLine_512();
+    // 更新 RLine_512 mask 的 blob 行标记（合并边缘检测结果后刷新）
+    int UpdateBlobHFlag_RLine_512();
+    // 获得 BQ标签 blob
+    int GetBlob_BQ();
+
+    // BQ标签 AI 检测
+    int AI_Detect_BQ();
 
     // 对AI mask图片进行 结果处理
     int AIMaskDet();
@@ -211,6 +223,8 @@ private:
     int AI_Detect_QX();
     // 127cell
     int AI_Detect_127Cell();
+    // 弱线检测-512模型
+    int AI_Detect_RLine_512();
     // 缺陷分类
     int AI_QX_Class_Thread();
 
@@ -237,12 +251,18 @@ private:
     // 临时绘制结果
     int DrawResult_Step_1();
 
+    // 绘制并保存结果图开关（base 配置 bDrawAndSave，替代 SAVE_RESIMG 宏）
+    bool NeedDrawAndSave();
+
     // 参数缺陷类型转 结果参数类型
     int AIClassTypeToConfigType(int nAIQXType, cv::Rect qx_roi);
 
     // 更新成像精度
     int UpdateImgageScale();
 
+    // 分类完成后按 bbox 完全相等去重（同一缺陷被多个 mask 来源各提取一次）
+    int DedupBlob_ByBBox();
+
     // 把blob 汇总成  检测结果。
     int BLobToDetResult();
     // 二次计算缺陷面积

AlgorithmModule/include/ImgCheckBase.h

@@ -2,7 +2,7 @@
 #define ImgCheckBase_H_
 #include <string>
 #include <memory>
-#define ALL_INTERFACE_VERSION 6
+#define ALL_INTERFACE_VERSION 7
 #define MAX_GPU_NUM 4
 
 enum CHECK_THREAD_RUN_STATUS

AlgorithmModule/include/ImgCheckConfig.h

@@ -18,7 +18,7 @@
 #include <unordered_map> // unordered_map
 #define RESULT_VERSION 31
 
-#define MAX_BLOB_NUM 200
+#define MAX_BLOB_NUM 1000
 
 // 输入模型图片尺寸
 #define SRCIMG_WIDTH 14200

AlgorithmModule/include/QX_Merge_Analysis.h

@@ -8,6 +8,8 @@
 #include <mutex>
 #include <thread>
 #include <condition_variable>
+#include <unordered_map>
+#include <vector>
 #include "CheckErrorCodeDefine.hpp"
 #include "CheckConfigDefine.h"
 #include "QX_Analysis.h"
@@ -68,6 +70,90 @@ public:
         }
     };
 
+    struct AD_Channel_BD_Info_
+    {
+        cv::Rect roi;
+        cv::Point2f location_product_mm;
+        vector<string> s_name_2;
+        vector<string> s_name_3;
+        int s_type; 
+        AD_Channel_BD_Info_()
+        {
+            roi = cv::Rect(0, 0, 0, 0);
+            location_product_mm = cv::Point2f(0, 0);
+            s_name_2.clear();
+            s_name_3.clear();
+            s_type = 0;
+        }
+        std::string GetInfo()
+        {
+            char buffer[256];
+            snprintf(buffer, sizeof(buffer), "roi:[%d,%d,%d,%d]  location [%f %f] s_type:%d",
+                    roi.x, roi.y, roi.width, roi.height, location_product_mm.x, location_product_mm.y, s_type);
+            std::string str = buffer;
+            for(int i = 0; i < s_name_2.size(); i++)
+            {
+                str += " s_name_2[" + s_name_2[i] + "]";
+            }
+            for(int i = 0; i < s_name_3.size(); i++)
+            {
+                str += " s_name_3[" + s_name_3[i] + "]";
+            }
+            return str;
+        }
+    };
+
+    struct AD_Channel_Stat_
+    {
+        int num_1s;
+        int num_2s;
+        int num_3s;
+        int num_2s_s;
+        int num_2s_m;
+        int num_3s_s;
+        int num_3s_m;
+        int product_AD_num;
+        float min_dis;
+        AD_Channel_Stat_()
+        {
+            num_1s = 0;
+            num_2s = 0;
+            num_3s = 0;
+            num_2s_s = 0;
+            num_2s_m = 0;
+            num_3s_s = 0;
+            num_3s_m = 0;
+            product_AD_num = 0;
+            min_dis = 9999999999;
+        }
+    };
+
+    // 关联通道缺陷信息(用于判断关联通道是否存在脏污(dirty)重叠缺陷)
+    struct Correlated_Qx_Info_
+    {
+        std::string channel_name; // 缺陷所属通道
+        std::string camera_name;  // 缺陷所属相机
+        int config_qx_type;       // 缺陷配置类型 CONFIG_QX_NAME_
+        int result_type;          // 缺陷结果类型(仅用于日志/追溯)
+        cv::Rect product_roi;     // 产品坐标系下的缺陷ROI
+        Correlated_Qx_Info_()
+        {
+            channel_name = "";
+            camera_name = "";
+            config_qx_type = -1;
+            result_type = -1;
+            product_roi = cv::Rect(0, 0, 0, 0);
+        }
+        std::string GetInfo()
+        {
+            char buffer[256];
+            snprintf(buffer, sizeof(buffer), "cam:%s channel:%s cfgType:%d resType:%d roi:[%d,%d,%d,%d]",
+                     camera_name.c_str(), channel_name.c_str(), config_qx_type, result_type,
+                     product_roi.x, product_roi.y, product_roi.width, product_roi.height);
+            return std::string(buffer);
+        }
+    };
+
 public:
     QX_Merge_Analysis();
     ~QX_Merge_Analysis();
@@ -88,6 +174,19 @@ public:
     bool Idx(int qxidx);
     bool AddQxInfo(int qxidx, QX_Info qxinfo, std::shared_ptr<DetLog> plog, std::shared_ptr<DetLog> pchannelLog);
 
+    // 登记关联通道(非主/次通道)的过滤源缺陷信息(类型由 correlated_types 配置, 如脏污/暗点)，供 AD 分析判断重叠缺陷
+    bool AddCorrelatedQxInfo(Correlated_Qx_Info_ qxinfo, std::shared_ptr<DetLog> plog);
+
+    // 判断缺陷类型是否为配置的关联过滤源类型(analysis_bind.correlated_types, 如 "Dirty_L0"/"Dirty_L1"/"Dirty_L2"/"AD")
+    // 配置为空时默认三种脏污; 登记与过滤两处共用此判定
+    bool IsCorrelatedSourceConfigType(int nconfigType);
+
+    // 关联通道过滤源重叠判断(供外部调用)：当前缺陷ROI是否与关联通道过滤源缺陷(脏污/暗点等)重叠
+    bool IsDirtyOverlapInCorrelatedChannel(const std::string &cur_channel_name,
+                                           const std::string &cur_camera_name,
+                                           const cv::Rect &qx_product_roi,
+                                           std::shared_ptr<DetLog> plog);
+
     bool setSendStatus(std::string cameraName, RUN_STATUS_ status);
 
     bool waitComplete(int timeoutMs = 15000);
@@ -115,7 +214,28 @@ private:
     int Judge_OK_Config();
     int Judge_NG_Config();
 
-    int Analysis_AD(ALL_Qx_DataList *pALLTypeqxList, int qx_idx);
+    // pAllowChannels 非空时仅列表内通道参与分析(主通道未配置时的回退: 仅次通道)；
+    // nullptr=不过滤；非空但为空列表=不做任何通道分析(直接返回)
+    int Analysis_AD(ALL_Qx_DataList *pALLTypeqxList, int qx_idx, const std::vector<std::string> *pAllowChannels = nullptr);
+    int Analysis_AD_Bind(ALL_Qx_DataList *pALLTypeqxList, int qx_idx);
+
+    // 绑定分析的主通道是否未配置(2s 或 3s 的主通道列表为空)；
+    // 只按配置判断，与本次产品是否检出缺陷无关(良品主通道本来就没暗点，按缺陷判断会频繁误回退)
+    bool IsADBindMainChannelConfigEmpty();
+    // 绑定分析的次通道列表(second_channel_2s + second_channel_3s, 去重并过滤空名)
+    std::vector<std::string> GetADBindSecondChannels();
+
+    // 关联过滤源过滤：移除与关联通道过滤源缺陷重叠的缺陷，使其不参与后续数量/距离/S分析
+    int FilterCorrelatedDirtyQx(ALL_Qx_DataList *pALLTypeqxList);
+
+    // 判断当前缺陷是否与关联通道(correlated_channels)中的过滤源缺陷(correlated_types)重叠
+    bool IsOverlapDirtyInCorrelatedChannel(const std::vector<std::string> &correlated_channels,
+                                           const std::string &cur_channel_name,
+                                           const std::string &cur_camera_name,
+                                           const cv::Rect &qx_product_roi,
+                                           int roi_expand,
+                                           float iou_threshold,
+                                           std::shared_ptr<DetLog> plog);
 
     ChannelCheckFunction *GetChannelFuntion(std::string strChannelName);
 
@@ -140,6 +260,13 @@ private:
 
     std::vector<AD_Channel_Info_> AD_list;
 
+    std::unordered_map<string, vector<AD_Channel_BD_Info_>> AD_list_bind;
+
+    // 关联通道缺陷缓存：channel_name -> 缺陷列表(产品坐标系)
+    std::unordered_map<std::string, std::vector<Correlated_Qx_Info_>> m_correlatedQxMap;
+    std::mutex mtx_correlatedQx;
+
+
 private:
 };
 

AlgorithmModule/src/AI_Edge_Algin.cpp

@@ -123,7 +123,7 @@ int AI_Edge_Algin::Detect(const cv::Mat &img, DetConfig *pDetConfig, std::shared
             return 7;
         }
 
-        if (max_contour_index >= 0)
+        if (max_contour_index >= 0&& contours.size() > 0)
         {
             result_roi = boundingRect(contours[max_contour_index]);
         }

AlgorithmModule/src/AI_Edge_QX_Det.cpp

@@ -122,18 +122,18 @@ std::string AI_Edge_QX_Det::GetImgSaveName(int flag)
 
     return str_root;
 }
-int AI_Edge_QX_Det::Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDetConfig, std::shared_ptr<DetLog> pdetlog)
+int AI_Edge_QX_Det::Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDetConfig, std::shared_ptr<DetLog> pdetlog, cv::Mat *pDetmask2)
 {
     static int n_count = 0;
     int nRuntype = 0; // 0:检测 1:测试。。。。。
     m_pdetlog = pdetlog;
     if (pDetConfig->ninstruct == 767)
     {
-        // nRuntype = 1;
+        nRuntype = 1;
     }
     else if (pDetConfig->ninstruct == 768)
     {
-        // nRuntype = 2;
+        nRuntype = 2;
     }
 
     bool bdege = false;
@@ -169,6 +169,14 @@ int AI_Edge_QX_Det::Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDet
 
     m_pdetlog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, "AI_Edge_QX_Det ", "nruntype  %d  %d\n", pDetConfig->pdege_Det_config->bsave_all, pDetConfig->pdege_Det_config->bsave_qx);
 
+    // 第二个合并目标（如弱线检测mask）：尺寸与主检测mask一致时才参与合并
+    bool bMergeDetmask2 = false;
+    if (pDetmask2 != nullptr && !pDetmask2->empty() && pDetmask2->size() == detmask.size())
+    {
+        bMergeDetmask2 = true;
+        m_pdetlog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, "AI_Edge_QX_Det ", "extra detmask merge on  %s", pDetConfig->strCamChannel.c_str());
+    }
+
     if (bdege || nRuntype != 0)
     {
         showImg = img.clone();
@@ -206,6 +214,16 @@ int AI_Edge_QX_Det::Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDet
 
             continue;
         }
+        if (bdege || nRuntype != 0)
+        {
+            std::string strSaveDir = "/home/aidlux/BOE/FOG/AI_Edge_Detect/";
+            CheckUtil::CreateDir(strSaveDir);
+            std::string roi_ext = "[" + std::to_string(r.x) + "," + std::to_string(r.y) + "]";
+            std::string strSavePath_in = strSaveDir + pDetConfig->strCamChannel + "_" + roi_ext + "_edge_in.png";
+            std::string strSavePath_out = strSaveDir + pDetConfig->strCamChannel + "_" + roi_ext + "_edge_out.png";
+            cv::imwrite(strSavePath_in, detimg);
+            cv::imwrite(strSavePath_out, outMask);
+        }
         bool isAllZero = true;
         if (cv::countNonZero(outMask) > 0)
         {
@@ -295,14 +313,27 @@ int AI_Edge_QX_Det::Detect(const cv::Mat &img, cv::Mat &detmask, DetConfig *pDet
             if (pDetConfig->pdege_Det_config->type == Function_Edge_Det::Edge_type_Add)
             {
                 outMask(smallImg_roi).copyTo(detmask(mask_roi), outMask(smallImg_roi));
+                // 第二个合并目标（如弱线检测mask）：与主检测mask相同规则合并
+                if (bMergeDetmask2)
+                {
+                    outMask(smallImg_roi).copyTo((*pDetmask2)(mask_roi), outMask(smallImg_roi));
+                }
             }
             else if (pDetConfig->pdege_Det_config->type == Function_Edge_Det::Edge_type_And)
             {
                 cv::bitwise_and(detmask(mask_roi), outMask(smallImg_roi), detmask(mask_roi));
+                if (bMergeDetmask2)
+                {
+                    cv::bitwise_and((*pDetmask2)(mask_roi), outMask(smallImg_roi), (*pDetmask2)(mask_roi));
+                }
             }
             else
             {
                 outMask(smallImg_roi).copyTo(detmask(mask_roi));
+                if (bMergeDetmask2)
+                {
+                    outMask(smallImg_roi).copyTo((*pDetmask2)(mask_roi));
+                }
             }
         }
     }

AlgorithmModule/src/CameraCheckAnalysisy.cpp

@@ -25,6 +25,7 @@
 #include <vector>
 #include <mutex>
 #include <algorithm>
+#include <ctime>
 
 static std::mutex g_ngImgSave_mutex;
 static std::vector<std::future<void>> g_ngImgSave_futures;
@@ -340,6 +341,26 @@ int CameraCheckAnalysisy::ResultParamJudge()
 		int re = m_imageResultJudge.ResultJudge(m_pCheck_Result->ImageALLDetResultList.at(imgidx));
 		long t2 = CheckUtil::getcurTime();
 	}
+
+	// 盲失联合判断：主通道(ms_main_channel)检出的盲失若在过滤通道(ms_filter_channels)同位置存在缺陷(不限制类型)，则不算盲失(撤销NG)
+	{
+		int nMsFilterNum = m_imageResultJudge.FilterMsBlindNG(m_pCheck_Result);
+		if (nMsFilterNum > 0)
+		{
+			m_pdetlog->AddCheckstr(PrintLevel_1, "ResultParamJudge", "盲失联合判断: 过滤通道同位置存在缺陷, 过滤盲失NG %d 个", nMsFilterNum);
+		}
+	}
+
+	// cell黑白点联合判断：主通道(cellbw_main_channel)检出的cell黑白点若在过滤通道(cellbw_filter_channels)同位置
+	// 存在缺陷(不限制类型)，则不算cell黑白点(撤销NG)
+	{
+		int nCellBwFilterNum = m_imageResultJudge.FilterCellBwNG(m_pCheck_Result);
+		if (nCellBwFilterNum > 0)
+		{
+			m_pdetlog->AddCheckstr(PrintLevel_1, "ResultParamJudge", "cell黑白点联合判断: 过滤通道同位置存在缺陷, 过滤NG %d 个", nCellBwFilterNum);
+		}
+	}
+
 	m_pQX_Merge_Analysis->setSendStatus(m_strcameraName, QX_Merge_Analysis::RUN_STATUS_READY);
 	m_pdetlog->AddCheckstr(PrintLevel_1, "ResultParamJudge", "===============QX_Merge_Analysis===start  %s", CheckUtil::getCurTimeHMS().c_str());
 
@@ -360,9 +381,10 @@ int CameraCheckAnalysisy::ResultParamJudge()
 		long t1 = CheckUtil::getcurTime();
 		int re = m_imageResultJudge.MergeResult(m_pCheck_Result->ImageALLDetResultList.at(imgidx), ptemre);
 		long t2 = CheckUtil::getcurTime();
-		#ifdef AI_Time
+		if (m_pbaseCheckFunction != NULL && m_pbaseCheckFunction->bDrawAndSave)
+		{
 			m_imageResultJudge.DrawResult(m_pCheck_Result->ImageALLDetResultList.at(imgidx));
-		#endif
+		}
 		long t3 = CheckUtil::getcurTime();
 		// m_pdetlog->AddCheckstr(PrintLevel_1, "ResultParamJudge", " det time %ld ms judge %ld  draw %ld", t3 - t1, t2 - t1, t3 - t2);
 
@@ -695,7 +717,13 @@ int CameraCheckAnalysisy::preDet_BQ(const cv::Mat &L255CutImg, std::shared_ptr<I
 {
 	long ts = CheckUtil::getcurTime();
 	m_pCheck_Result->cameraBaseResult->pBQ_Result = std::make_shared<BQ_Result>();
-	m_pCheck_Result->cameraBaseResult->pBQ_Result->bShield_BQ = m_AnalysisyConfig.commonCheckConfig.baseConfig.bShield_BQ;
+
+	ChannelCheckFunction *pChannelFuntion_BQ = GetChannelFuntion(L255->strChannel);
+	if (pChannelFuntion_BQ)
+	{
+		m_pCheck_Result->cameraBaseResult->pBQ_Result->bBQ_AI_Det = pChannelFuntion_BQ->function.f_Det_BQ.bBQ_AI_Det;
+		m_pCheck_Result->cameraBaseResult->pBQ_Result->bsaveAll = pChannelFuntion_BQ->function.f_Det_BQ.bsaveAll;
+	}
 
 	cv::Rect Det_CropRoi = m_pCheck_Result->cameraBaseResult->pEdgeDetResult->cutRoi;
 
@@ -716,25 +744,36 @@ int CameraCheckAnalysisy::preDet_BQ(const cv::Mat &L255CutImg, std::shared_ptr<I
 		return re;
 	}
 
-	// 如果要屏蔽标签
-	if (m_pCheck_Result->cameraBaseResult->pBQ_Result->bShield_BQ)
+	// 默认屏蔽标签
 	{
-		m_pdetlog->AddCheckstr(PrintLevel_2, "preDet_BQ", "shieldBQ is Open");
 		cv::Mat detImg_mask = m_pCheck_Result->cameraBaseResult->pEdgeDetResult->shieldMask;
 		for (int i = 0; i < m_pCheck_Result->cameraBaseResult->pBQ_Result->pBQ_roiList.size(); i++)
 		{
 			cv::Rect boundingRect = m_pCheck_Result->cameraBaseResult->pBQ_Result->pBQ_roiList.at(i);
+			int BQ_expand = pChannelFuntion_BQ ? pChannelFuntion_BQ->function.f_Det_BQ.BQ_expand : 0;
+			boundingRect.x -= BQ_expand;
+			boundingRect.y -= BQ_expand;
+			boundingRect.width += 2 * BQ_expand;
+			boundingRect.height += 2 * BQ_expand;
+			boundingRect = boundingRect & cv::Rect(0, 0, detImg_mask.cols, detImg_mask.rows);
+			if (boundingRect.width > 0 && boundingRect.height > 0)
+			{
 				detImg_mask(boundingRect).setTo(255);
+				m_pCheck_Result->cameraBaseResult->pBQ_Result->pBQ_expandedRoiList.push_back(boundingRect);
+
+				// 保存裁剪图像供独立BQ AI检测
+				if (pChannelFuntion_BQ && pChannelFuntion_BQ->function.f_Det_BQ.bBQ_AI_Det)
+				{
+					cv::Mat bqCrop = L255CutImg(boundingRect).clone();
+					m_pCheck_Result->cameraBaseResult->pBQ_Result->pBQ_cropImages.push_back(bqCrop);
+				}
+			}
 		}
 		if (L255->result->in_shareImage->bDebugsaveImg)
 		{
 			cv::imwrite("sheildImg.png", m_pCheck_Result->cameraBaseResult->pEdgeDetResult->shieldMask);
 		}
 	}
-	else
-	{
-		m_pdetlog->AddCheckstr(PrintLevel_2, "preDet_BQ", "shieldBQ is false");
-	}
 	std::shared_ptr<BQ_Result> pBQ_Result = m_pCheck_Result->cameraBaseResult->pBQ_Result; // BQ 检测结果
 	pBQ_Result->BQ_centerPoint.erase(pBQ_Result->BQ_centerPoint.begin(), pBQ_Result->BQ_centerPoint.end());
 	for (int i = 0; i < pBQ_Result->pBQ_roiList.size(); i++)
@@ -1235,43 +1274,55 @@ int CameraCheckAnalysisy::ngImgSave() {
     auto result_copy = m_pCheck_Result;
     std::string save_path_root = "/home/aidlux/BOE/FOG/cloud/";
 
-    auto task = std::async(std::launch::async, [result_copy, save_path_root]() {
+    // 生成当天日期字符串和当前小时字符串，用于按天/小时组织存储目录
+    time_t now = time(nullptr);
+    struct tm tm_now;
+    localtime_r(&now, &tm_now);
+    char date_buf[16];
+    strftime(date_buf, sizeof(date_buf), "%Y-%m-%d", &tm_now);
+    std::string date_str(date_buf);
+    char hour_buf[4];
+    strftime(hour_buf, sizeof(hour_buf), "%H", &tm_now);
+    std::string hour_str(hour_buf);
+
+    auto task = std::async(std::launch::async, [result_copy, save_path_root, date_str, hour_str]() {
         try {
             if (!result_copy) return;
 
             // ---------- 目录清理（加锁防止并发删除）----------
+            // 只保留最近 2 天的数据（今天 + 昨天），删除更早的日期目录
             {
                 static std::mutex cleanup_mutex;  // 静态锁，所有实例共享
                 std::lock_guard<std::mutex> lock(cleanup_mutex);
 
+                // 计算 1 天前的日期作为清理截止线（保留今天和昨天共 2 天）
+                time_t now_cleanup = time(nullptr);
+                time_t cutoff_time = now_cleanup - 1 * 24 * 3600;
+                struct tm tm_cutoff;
+                localtime_r(&cutoff_time, &tm_cutoff);
+                char cutoff_buf[16];
+                strftime(cutoff_buf, sizeof(cutoff_buf), "%Y-%m-%d", &tm_cutoff);
+                std::string cutoff_str(cutoff_buf);
+
                 DIR* dir = opendir(save_path_root.c_str());
                 if (dir) {
-                    std::vector<std::pair<std::string, time_t>> dirs;
                     struct dirent* entry;
                     while ((entry = readdir(dir)) != nullptr) {
                         std::string name = entry->d_name;
                         if (name == "." || name == "..") continue;
+                        // 检查是否为日期格式目录 YYYY-MM-DD
+                        if (name.length() == 10 && name[4] == '-' && name[7] == '-') {
+                            // ISO 日期格式可按字典序直接比较
+                            if (name < cutoff_str) {
                                 std::string full_path = save_path_root + name;
                                 struct stat st;
                                 if (stat(full_path.c_str(), &st) == 0 && S_ISDIR(st.st_mode)) {
-                            struct stat lst;
-                            if (lstat(full_path.c_str(), &lst) == 0 && !S_ISLNK(lst.st_mode)) {
-                                dirs.emplace_back(full_path, st.st_mtime);
-                            }
+                                    CheckUtil::DeleteDir(full_path);
                                 }
                             }
-                    closedir(dir);
-
-                    if (dirs.size() > 100) {
-                        std::sort(dirs.begin(), dirs.end(),
-                                  [](auto& a, auto& b) { return a.second < b.second; });
-                        const std::string& oldest_dir = dirs.front().first;
-                        if (!oldest_dir.empty() && oldest_dir != save_path_root) {
-                            CheckUtil::DeleteDir(oldest_dir);
                         }
                     }
-                } else {
-                    cerr << ("ERROR: cannot open root directory for cleanup");
+                    closedir(dir);
                 }
             }
 
@@ -1295,13 +1346,14 @@ int CameraCheckAnalysisy::ngImgSave() {
                 std::string safe_channel = SanitizePathComponent(raw_channel);
 
                 // 只保存NG结果且图像数据非空
-                if ((m_CheckResult_shareP->nresult == 1 || safe_channel == "L255") &&
-                    !m_CheckResult_shareP->resultMaskImg.empty() &&
-                    !m_CheckResult_shareP->resultimg.empty() &&
-                    !m_CheckResult_shareP->cutSrcimg.empty()) {
+                // if ((m_CheckResult_shareP->nresult == 1 || safe_channel == "L255") &&
+                //     !m_CheckResult_shareP->resultMaskImg.empty() &&
+                //     !m_CheckResult_shareP->resultimg.empty() &&
+                //     !m_CheckResult_shareP->cutSrcimg.empty()) 
+				{
 
-                    // 构建保存目录（使用净化后的组件）
-                    std::string save_dir = save_path_root + "/" + safe_product + "/" + safe_camera + "/";
+                    // 构建保存目录（使用净化后的组件，包含日期/小时子目录）
+                    std::string save_dir = save_path_root + date_str + "/" + hour_str + "/" + safe_product + "/" + safe_camera + "/";
 					if (CheckUtil::CreateDir(save_dir) != 0) {
                         cerr << ("Failed to create directory: " + save_dir);
                         continue;
@@ -1316,7 +1368,7 @@ int CameraCheckAnalysisy::ngImgSave() {
                         // cv::imwrite(save_dir + safe_channel + "_AIMask.png", mask_img);
                         // cv::imwrite(save_dir + safe_channel + "_Resultimg.png", result_img);
                         // cv::imwrite(save_dir + safe_channel + "_CutImg.png", cut_img);
-                        writeLog(save_dir+ safe_channel + "log", m_CheckResult_shareP->det_LogList);
+                        writeLog(save_dir+ safe_channel + "_log", m_CheckResult_shareP->det_LogList);
                     } catch (const cv::Exception& e) {
                         cerr << ("OpenCV exception: " + std::string(e.what()));
                     } catch (const std::exception& e) {

AlgorithmModule/src/CheckUtil.cpp

@@ -457,6 +457,32 @@ int CheckUtil::CalHj(const cv::Mat &img, const cv::Mat &mask, const cv::Mat &bac
 	return meanValue123[0];
 }
 
+int CheckUtil::CalHjWeighted(const cv::Mat &img, const cv::Mat &mask, int b_value, float power)
+{
+	cv::Mat imgf;
+	if (img.type() != CV_32F)
+		img.convertTo(imgf, CV_32F);
+	else
+		imgf = img;
+
+	cv::Mat absDiff = cv::abs(imgf - b_value);
+
+	cv::Mat weight;
+	cv::pow(absDiff, power, weight);
+
+	cv::Mat maskedDiff, maskedWeight;
+	absDiff.copyTo(maskedDiff, mask);
+	weight.copyTo(maskedWeight, mask);
+
+	cv::Scalar weightedSum = cv::sum(maskedDiff.mul(weight));
+	cv::Scalar weightSum = cv::sum(maskedWeight);
+
+	if (weightSum[0] < 1e-6)
+		return 0;
+
+	return static_cast<int>(weightedSum[0] / weightSum[0]);
+}
+
 float CheckUtil::CalRoi2RoiPre(cv::Rect rect1, cv::Rect rect2)
 {
 	// 计算交集区域

AlgorithmModule/src/ImageResultJudge.cpp

@@ -1,5 +1,7 @@
 
 #include "ImageResultJudge.h"
+#include <cmath>
+#include <algorithm>
 
 ImageResultJudge::ImageResultJudge(/* args */)
 {
@@ -96,6 +98,12 @@ cv::Rect ImageResultJudge::GetCutRoi(cv::Rect &roi, const cv::Mat &img)
 
 	return cutroi;
 }
+bool ImageResultJudge::NeedDrawAndSave()
+{
+	// base 层“是否存图绘制”开关：绘制并保存结果图（运行时替代 SAVE_RESIMG 宏）
+	return m_pbaseCheckFunction != NULL && m_pbaseCheckFunction->bDrawAndSave;
+}
+
 int ImageResultJudge::GetAIDetImg(std::shared_ptr<ImageAllResult> pImageResult, cv::Point pcenter, cv::Mat &AI_InImg, cv::Mat &AI_OutImg)
 {
 
@@ -222,6 +230,25 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 
 		QX_info->result = QX_RESULT_TYPE_NoJduge;
 
+		// 关联通道分析：登记关联过滤源缺陷位置(类型由 analysis_bind.correlated_types 配置, 如脏污/暗点)，
+		// 供 AD 分析判断关联通道是否存在重叠缺陷。
+		// 登记只依赖缺陷类型与位置，与判定结果/是否绘制无关：
+		// 被分到 NoJudge(不判定/Draw)的缺陷也要登记(如 DP 上未开启 AD 判定的暗点)，否则无法过滤其它通道的重叠缺陷
+		if (m_pQX_Merge_Analysis->IsCorrelatedSourceConfigType(config_qx_type))
+		{
+			QX_Merge_Analysis::Correlated_Qx_Info_ corr_qx;
+			corr_qx.channel_name = pImageResult->strChannel;
+			corr_qx.camera_name = pImageResult->cameraBaseResult->strCameraName;
+			corr_qx.config_qx_type = config_qx_type;
+			corr_qx.result_type = ConfigTypeToResultType(config_qx_type);
+			corr_qx.product_roi = roi;
+			corr_qx.product_roi.x += fproduct_offx_piexl;
+
+			bool bAddCorr = m_pQX_Merge_Analysis->AddCorrelatedQxInfo(corr_qx, pQxLog);
+			pQxLog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, " Use  Num And Dis Judge ", "qx %s ,Add Correlated %s   %s",
+								qx_name.c_str(), BOOL_TO_STR_Error(bAddCorr), corr_qx.GetInfo().c_str());
+		}
+
 		// 在其他位置已经判断了 不进入分析。直接绘制。
 		if (qx_status == QX_Stauts_Draw)
 		{
@@ -327,6 +354,7 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 						float hj = pParam->paramArr[j].hj;
 						float Len = pParam->paramArr[j].length;
 						float md = pParam->paramArr[j].density;
+						float area_max = pParam->paramArr[j].area_max;
 
 						Judge_Status = true;
 
@@ -385,6 +413,31 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 												BOOL_TO_STR(flen >= Len), flen, BOOL_TO_ThanLess(flen > Len), Len);
 						}
 
+						if (config_qx_type == CONFIG_QX_NAME_POL_Cell && area_max > 0)
+						{
+							if (detArea > area_max)
+							{
+								nerrortype = 0;
+								result = true;
+								bNG_Status = false;
+								bYS_Status = false;
+							}
+							pQxLog->AddCheckstr(PrintLevel_4, DET_LOG_LEVEL_3, "PolCell_Area_Max", "%s -> %f %s %f ",
+												BOOL_TO_STR(detArea > area_max), detArea, BOOL_TO_ThanLess(detArea > area_max), area_max);
+						}
+						// AD根据配置检测：未配置通道的AD缺陷不判NG
+						if (config_qx_type == CONFIG_QX_NAME_AD && m_pbaseCheckFunction && ict == ANALYSIS_TYPE_TF &&
+							std::find(m_pbaseCheckFunction->ad_check.analysis_sf.region_sf.begin(),
+									  m_pbaseCheckFunction->ad_check.analysis_sf.region_sf.end(),
+									  strChannel) == m_pbaseCheckFunction->ad_check.analysis_sf.region_sf.end())
+						{
+							nerrortype = 0;
+							result = true;
+							bNG_Status = false;
+							bYS_Status = false;
+							pQxLog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, "Param Judge", "AD channel %s not configured in region_sf", strChannel.c_str());
+						}
+
 						if (!result)
 						{
 							break;
@@ -415,36 +468,37 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 				Judge_Status = true;
 				checkFlage = ANALYSIS_TYPE_TF;
 				nerrortype = 1;
+				bNG_Status = true;
 			}
 		}
 		// 暗点 3S 分析  对 好品 或者 是 疑是的缺陷进行3S分析。S数量分析。 3S的暗点直接NG.
-		// if ((nerrortype == 0 || checkFlage != ANALYSIS_TYPE_TF) &&
-		// 	config_qx_type == CONFIG_QX_NAME_AD && pFuntion &&
-		// 	pFuntion->function.f_AD_Check.bOpen &&
-		// 	pFuntion->function.f_AD_Check.analysis_s.bOpen &&
-		// 	pFuntion->function.f_AD_Check.analysis_s.NG_3s && bpreSucc)
 		if ((nerrortype == 0 || checkFlage != ANALYSIS_TYPE_TF) &&
-			config_qx_type == CONFIG_QX_NAME_AD && m_pbaseCheckFunction &&
+			config_qx_type == CONFIG_QX_NAME_AD && pFuntion != NULL && m_pbaseCheckFunction &&
 			m_pbaseCheckFunction->ad_check.bOpen &&
 			m_pbaseCheckFunction->ad_check.analysis_s.bOpen &&
 			m_pbaseCheckFunction->ad_check.analysis_s.NG_3s && bpreSucc)
 		{
+			bool b3sChannel = false;
+			if (std::find(m_pbaseCheckFunction->ad_check.analysis_sf.s_sf.begin(), m_pbaseCheckFunction->ad_check.analysis_sf.s_sf.end(), strChannel) != m_pbaseCheckFunction->ad_check.analysis_sf.s_sf.end())
+			{
+				b3sChannel = true;
+			}
 			Judge_Status = true;
 			int det_s_Vale = 0;
-			if (JudgArea >= pFuntion->function.f_AD_Check.S_standard_3s.area &&
-				flen >= pFuntion->function.f_AD_Check.S_standard_3s.len)
+			if (JudgArea >= m_pbaseCheckFunction->ad_check.S_standard_3s.area &&
+				flen >= m_pbaseCheckFunction->ad_check.S_standard_3s.len && b3sChannel)
 			{
 				nerrortype = 1;
 				checkFlage = ANALYSIS_TYPE_TF;
 				bNG_Status = true;
 
 				pQxLog->AddCheckstr(PrintLevel_3, DET_LOG_LEVEL_3, "info", " AD 3S result NG, det area %f len %f >=3s param area %f len %f  ",
-									JudgArea, flen, pFuntion->function.f_AD_Check.S_standard_3s.area, pFuntion->function.f_AD_Check.S_standard_3s.len);
+									JudgArea, flen, m_pbaseCheckFunction->ad_check.S_standard_3s.area, m_pbaseCheckFunction->ad_check.S_standard_3s.len);
 			}
 			else
 			{
 				pQxLog->AddCheckstr(PrintLevel_3, DET_LOG_LEVEL_3, "info", " AD 3S result OK, det area %f len %f < 3s param area %f len %f  ",
-									JudgArea, flen, pFuntion->function.f_AD_Check.S_standard_3s.area, pFuntion->function.f_AD_Check.S_standard_3s.len);
+									JudgArea, flen, m_pbaseCheckFunction->ad_check.S_standard_3s.area, m_pbaseCheckFunction->ad_check.S_standard_3s.len);
 			}
 		}
 
@@ -479,7 +533,7 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 		{
 
 			// 目前为止判断是好品
-			if (!bNG_Status && pFuntion->function.f_Big_QX.bOpen)
+			if (!bNG_Status && pFuntion != NULL && pFuntion->function.f_Big_QX.bOpen)
 			{
 				Judge_Status = true;
 				bool berror = false;
@@ -503,7 +557,7 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 			}
 
 			// 总面积判断
-			if (!bNG_Status && pFuntion->function.f_Big_QX.bOpen)
+			if (!bNG_Status && pFuntion != NULL && pFuntion->function.f_Big_QX.bOpen)
 			{
 				if (nSumBLobNUm <= pFuntion->function.f_Big_QX.Sum_blob_Num)
 				{
@@ -550,6 +604,17 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 			pQxLog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, " big qx ", "qx %s ,result = %s", qx_name.c_str(), resultType.c_str());
 		}
 
+		// 对于 AD_YX 类型，面积小于设定值的 降级为NoJudge
+		if (config_qx_type == CONFIG_QX_NAME_AD_YX && bYS_Status && pFuntion != NULL &&
+			pFuntion->function.f_YXDet.bOpen && JudgArea < pFuntion->function.f_YXDet.minArea)
+		{
+			bYS_Status = false;
+			QX_info->result = QX_RESULT_TYPE_OK;
+			QX_info->result_name = QX_RESULT_TYPE_Names[QX_info->result];
+			resultType = QX_info->result_name;
+			pQxLog->AddCheckstr(PrintLevel_2, DET_LOG_LEVEL_3, "AD_YX downgrade", "area %f < %d, YS -> OK", JudgArea, pFuntion->function.f_YXDet.minArea);
+		}
+
 		if (bNG_Status)
 		{
 			m_CheckResult_shareP->nresult = 1;
@@ -571,7 +636,8 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 			// CheckUtil::printROI(CutRoi, "CutRoi");
 			cv::Size sz = cv::Size(QX_SAMLLIMG_WIDTH, QX_SAMLLIMG_HEIGHT);
 
-			#ifdef AI_Time
+			if (NeedDrawAndSave())
+			{
 				if (pImageResult->detImg.channels() == 1)
 				{
 					cv::cvtColor(pImageResult->detImg(CutRoi), tem.srcImg, cv::COLOR_GRAY2BGR);
@@ -581,7 +647,7 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 					tem.srcImg = pImageResult->detImg(CutRoi).clone();
 				}
 				cv::resize(tem.srcImg, tem.resizeImg, sz);
-			#endif
+			}
 
 			int nqx_type = ConfigTypeToResultType(config_qx_type);
 			tem.type = nqx_type;
@@ -626,7 +692,10 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 					m_CheckResult_shareP->nresult = nqx_type;
 				}
 
+				if (NeedDrawAndSave())
+				{
 					GetAIDetImg(pImageResult, pCenter, tem.AI_in_Img, tem.AI_out_img);
+				}
 				m_CheckResult_shareP->qxImageResult.push_back(tem);
 			}
 			else
@@ -637,13 +706,15 @@ int ImageResultJudge::ResultJudge(std::shared_ptr<ImageAllResult> pImageResult)
 					m_CheckResult_shareP->nYS_result = nqx_type;
 				}
 
+				if (NeedDrawAndSave())
+				{
 					GetAIDetImg(pImageResult, pCenter, tem.AI_in_Img, tem.AI_out_img);
+				}
 
 				m_CheckResult_shareP->YS_ImageResult.push_back(tem);
 			}
 
 			pQxLog->AddCheckstr(PrintLevel_4, DET_LOG_LEVEL_3, " result ", " name %s ---code %s", tem.strTypeName.c_str(), tem.qx_Code.c_str());
-
 		}
 
 		// 对于大面积 的zara和 异显NG ，可以不接着分析后续的数据了。应为后续估计有很多错误类别。
@@ -747,6 +818,573 @@ int ImageResultJudge::UpdateMergedet(std::shared_ptr<CameraResult> pCheck_Result
 	return 0;
 }
 
+/**
+ * @brief 关联过滤源过滤：撤销与关联通道过滤源缺陷重叠的暗点NG结果(含单图判定的单点NG)
+ *        被过滤的暗点不再计入NG
+ * @param pImageResult 当前检测图片结果
+ * @return int 被过滤的缺陷个数
+ */
+int ImageResultJudge::FilterCorrelatedDirtyNG(std::shared_ptr<ImageAllResult> pImageResult)
+{
+	if (pImageResult == nullptr || pImageResult->cameraBaseResult == nullptr ||
+		m_pbaseCheckFunction == nullptr || m_pQX_Merge_Analysis == nullptr || m_pAnalysisyConfig == nullptr)
+	{
+		return 0;
+	}
+	// 未配置关联通道(NONE/空)时不需要判断
+	std::vector<std::string> correlated_channels = m_pbaseCheckFunction->ad_check.analysis_bind.correlated_channels;
+	if (correlated_channels.empty())
+	{
+		return 0;
+	}
+	std::shared_ptr<One_Image_CheckResult_> pDetResult = pImageResult->pDetResult; // 检测结果
+	std::shared_ptr<CheckResult> m_CheckResult_shareP = pImageResult->result;	   // 返回结果
+	if (pDetResult == nullptr || pDetResult->pQx_ErrorList == nullptr || m_CheckResult_shareP == nullptr)
+	{
+		return 0;
+	}
+
+	std::shared_ptr<DetLog> pImageLog = pImageResult->detlog;
+	float fproduct_offx_mm = m_pAnalysisyConfig->commonCheckConfig.baseConfig.fProduct_Off_X_mm;
+	int fproduct_offx_piexl = int(fproduct_offx_mm / m_fImgage_Scale_X);
+
+	int nFilterNum = 0;
+	for (int qxidx = 0; qxidx < (int)pDetResult->pQx_ErrorList->size(); qxidx++)
+	{
+		QX_ERROR_INFO_ *QX_info = &pDetResult->pQx_ErrorList->at(qxidx);
+		// 只处理 暗点(AD) 的 NG 结果
+		if (QX_info->nconfig_qx_type != CONFIG_QX_NAME_AD || QX_info->result != QX_RESULT_TYPE_NG)
+		{
+			continue;
+		}
+
+		// 关联通道存在与当前缺陷重叠的过滤源缺陷时，过滤该暗点NG
+		cv::Rect product_roi = QX_info->roi;
+		product_roi.x += fproduct_offx_piexl;
+		if (!m_pQX_Merge_Analysis->IsDirtyOverlapInCorrelatedChannel(pImageResult->strChannel,
+																	 pImageResult->cameraBaseResult->strCameraName,
+																	 product_roi,
+																	 pImageLog))
+		{
+			continue;
+		}
+
+		// 撤销NG结果
+		UndoNGResult(pImageResult, qxidx, QX_info);
+
+		nFilterNum++;
+		if (pImageLog)
+		{
+			pImageLog->AddCheckstr(PrintLevel_1, DET_LOG_LEVEL_3, "DirtyCorrelatedFilter",
+								   "img %s cam %s blob %d qx %s roi %s : 关联通道存在重叠的过滤源缺陷, 过滤该暗点NG",
+								   pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+								   qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str());
+		}
+		if (m_pQX_Merge_Analysis->m_pMergedetlog)
+		{
+			m_pQX_Merge_Analysis->m_pMergedetlog->AddCheckstr(PrintLevel_1, 3, "DirtyCorrelatedFilter",
+															  "img %s cam %s blob %d qx %s roi %s : 关联通道存在重叠的过滤源缺陷, 过滤该暗点NG",
+															  pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+															  qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str());
+		}
+	}
+
+	return nFilterNum;
+}
+
+/**
+ * @brief 撤销单个缺陷的NG结果：恢复为OK，移除对应的NG小图并回退计数
+ * @param pImageResult 当前检测图片结果
+ * @param qxidx 缺陷在 pQx_ErrorList 中的下标
+ * @param QX_info 缺陷信息
+ * @return int 移除的NG小图个数
+ */
+int ImageResultJudge::UndoNGResult(std::shared_ptr<ImageAllResult> pImageResult, int qxidx, QX_ERROR_INFO_ *QX_info)
+{
+	if (pImageResult == nullptr || QX_info == nullptr)
+	{
+		return 0;
+	}
+	std::shared_ptr<CheckResult> m_CheckResult_shareP = pImageResult->result;
+	if (m_CheckResult_shareP == nullptr)
+	{
+		return 0;
+	}
+
+	// 撤销NG结果
+	QX_info->result = QX_RESULT_TYPE_OK;
+	QX_info->result_name = QX_RESULT_TYPE_Names[QX_info->result];
+	int nqx_type = ConfigTypeToResultType(QX_info->nconfig_qx_type);
+
+	// 移除该缺陷的NG小图
+	int nEraseNum = 0;
+	for (auto it = m_CheckResult_shareP->qxImageResult.begin(); it != m_CheckResult_shareP->qxImageResult.end();)
+	{
+		if (it->idx == qxidx)
+		{
+			it = m_CheckResult_shareP->qxImageResult.erase(it);
+			nEraseNum++;
+		}
+		else
+		{
+			++it;
+		}
+	}
+
+	// 计数回退
+	if (nqx_type >= 0 && nqx_type < ERROR_TYPE_COUNT)
+	{
+		m_CheckResult_shareP->defectResultList[nqx_type].num -= nEraseNum;
+		if (m_CheckResult_shareP->defectResultList[nqx_type].num < 0)
+		{
+			m_CheckResult_shareP->defectResultList[nqx_type].num = 0;
+		}
+	}
+	if (pImageResult->NG_num > 0)
+	{
+		pImageResult->NG_num--;
+	}
+	pImageResult->Ok_num++;
+	// 过滤后已无NG，则恢复图片结果
+	if (m_CheckResult_shareP->qxImageResult.empty())
+	{
+		m_CheckResult_shareP->nresult = ERROR_TYPE_OK;
+	}
+
+	return nEraseNum;
+}
+
+/**
+ * @brief 判断盲失(产品坐标系ROI)在配置的过滤通道是否存在同位置缺陷(不限制缺陷类型)
+ *        同位置判断方式参考暗点联合分析：当前ROI与过滤通道缺陷ROI双方均按 expand 扩张后计算 IOU
+ * @param imgResults 当前相机所有通道的检测结果
+ * @param cur_channel_name 盲失所属通道(主通道)
+ * @param qx_product_roi 盲失的产品坐标系ROI
+ * @param roi_expand 扩张像素(ms_expand)
+ * @param iou_threshold IOU阈值(ms_iou)
+ * @param plog 日志
+ * @return bool 过滤通道同位置存在缺陷返回true
+ */
+bool ImageResultJudge::IsOverlapQxInMsFilterChannel(const std::vector<std::shared_ptr<ImageAllResult>> &imgResults,
+													const std::string &cur_channel_name,
+													const cv::Rect &qx_product_roi,
+													int roi_expand, float iou_threshold,
+													std::shared_ptr<DetLog> plog)
+{
+	// 防御：ROI非法时不判断，避免无效计算与除零
+	if (qx_product_roi.width <= 0 || qx_product_roi.height <= 0)
+	{
+		return false;
+	}
+	if (m_pbaseCheckFunction == nullptr || m_pAnalysisyConfig == nullptr)
+	{
+		return false;
+	}
+	const std::vector<std::string> &filter_channels = m_pbaseCheckFunction->ms_check.ms_filter_channels;
+	if (filter_channels.empty())
+	{
+		return false;
+	}
+	float fproduct_offx_mm = m_pAnalysisyConfig->commonCheckConfig.baseConfig.fProduct_Off_X_mm;
+	int fproduct_offx_piexl = int(fproduct_offx_mm / m_fImgage_Scale_X);
+
+	// 与暗点联合分析保持一致：双方均按 expand 扩张后再计算 IOU
+	cv::Rect roi_expanded(qx_product_roi.x - roi_expand, qx_product_roi.y - roi_expand,
+						  qx_product_roi.width + roi_expand * 2, qx_product_roi.height + roi_expand * 2);
+
+	for (int imgidx = 0; imgidx < (int)imgResults.size(); imgidx++)
+	{
+		std::shared_ptr<ImageAllResult> pFilterImage = imgResults.at(imgidx);
+		if (pFilterImage == nullptr)
+		{
+			continue;
+		}
+		// 只与配置的过滤通道比较；当前通道不参与，避免自身重叠导致误过滤
+		if (pFilterImage->strChannel == cur_channel_name)
+		{
+			continue;
+		}
+		if (std::find(filter_channels.begin(), filter_channels.end(), pFilterImage->strChannel) == filter_channels.end())
+		{
+			continue;
+		}
+		std::shared_ptr<One_Image_CheckResult_> pFilterDetResult = pFilterImage->pDetResult;
+		if (pFilterDetResult == nullptr || pFilterDetResult->pQx_ErrorList == nullptr)
+		{
+			continue;
+		}
+
+		// 过滤通道不限制缺陷类型：任意类型缺陷都参与同位置比较，与具体判定结果无关
+		for (int j = 0; j < (int)pFilterDetResult->pQx_ErrorList->size(); j++)
+		{
+			const QX_ERROR_INFO_ &filter_qx = pFilterDetResult->pQx_ErrorList->at(j);
+			cv::Rect filter_roi = filter_qx.roi;
+			filter_roi.x += fproduct_offx_piexl;
+			cv::Rect filter_expanded(filter_roi.x - roi_expand, filter_roi.y - roi_expand,
+									 filter_roi.width + roi_expand * 2, filter_roi.height + roi_expand * 2);
+
+			float fiou = CheckUtil::CalIoU(roi_expanded, filter_expanded);
+			if (plog)
+			{
+				plog->AddCheckstr(PrintLevel_3, DET_LOG_LEVEL_3, "MsBlind",
+								  "main channel %s roi %s <-> filter channel %s qx %d roi %s expand %d iou %f",
+								  cur_channel_name.c_str(), CheckUtil::GetRectString(qx_product_roi).c_str(),
+								  pFilterImage->strChannel.c_str(), j,
+								  CheckUtil::GetRectString(filter_roi).c_str(), roi_expand, fiou);
+			}
+			// IOU 为 NaN(ROI 非法导致的除零)时比较结果为 false，不会误过滤
+			if (fiou > iou_threshold)
+			{
+				if (plog)
+				{
+					plog->AddCheckstr(PrintLevel_1, DET_LOG_LEVEL_3, "MsBlind",
+									  "main channel %s roi %s overlap filter channel %s qx %d roi %s expand %d iou %f > %f",
+									  cur_channel_name.c_str(), CheckUtil::GetRectString(qx_product_roi).c_str(),
+									  pFilterImage->strChannel.c_str(), j,
+									  CheckUtil::GetRectString(filter_roi).c_str(), roi_expand, fiou, iou_threshold);
+				}
+				return true;
+			}
+		}
+	}
+
+	return false;
+}
+
+/**
+ * @brief 盲失联合判断：主通道(ms_main_channel)检出的盲失(Other)需要和过滤通道(ms_filter_channels)同位置判断，
+ *        过滤通道同位置存在缺陷(不限制缺陷类型)时不算盲失(撤销NG)
+ *        同位置判断方式参考暗点联合分析：双方按 ms_expand 扩张后计算 IOU, 大于 ms_iou 认为同位置
+ * @param pCheck_Result 当前相机检测结果(包含该相机所有通道的检测结果)
+ * @return int 被过滤的盲失NG个数
+ */
+int ImageResultJudge::FilterMsBlindNG(std::shared_ptr<CameraResult> pCheck_Result)
+{
+	if (pCheck_Result == nullptr || m_pbaseCheckFunction == nullptr || m_pAnalysisyConfig == nullptr)
+	{
+		return 0;
+	}
+	Base_Function_MS_Check *pMsParam = &m_pbaseCheckFunction->ms_check;
+	// 未开启盲失分析 或 未配置主通道/过滤通道 时不需要判断
+	if (!pMsParam->bOpen || pMsParam->ms_main_channel.empty() || pMsParam->ms_filter_channels.empty())
+	{
+		return 0;
+	}
+
+	const int roi_expand = (int)(round)(pMsParam->MS_Expand);
+	const float iou_threshold = pMsParam->MS_IOU;
+	float fproduct_offx_mm = m_pAnalysisyConfig->commonCheckConfig.baseConfig.fProduct_Off_X_mm;
+	int fproduct_offx_piexl = int(fproduct_offx_mm / m_fImgage_Scale_X);
+
+	int nFilterNum = 0;
+	for (int imgidx = 0; imgidx < (int)pCheck_Result->ImageALLDetResultList.size(); imgidx++)
+	{
+		std::shared_ptr<ImageAllResult> pImageResult = pCheck_Result->ImageALLDetResultList.at(imgidx);
+		if (pImageResult == nullptr)
+		{
+			continue;
+		}
+		// 只处理配置的主通道
+		if (std::find(pMsParam->ms_main_channel.begin(), pMsParam->ms_main_channel.end(), pImageResult->strChannel) == pMsParam->ms_main_channel.end())
+		{
+			continue;
+		}
+		std::shared_ptr<One_Image_CheckResult_> pDetResult = pImageResult->pDetResult; // 检测结果
+		if (pDetResult == nullptr || pDetResult->pQx_ErrorList == nullptr || pImageResult->result == nullptr)
+		{
+			continue;
+		}
+		std::shared_ptr<DetLog> pImageLog = pImageResult->detlog;
+
+		for (int qxidx = 0; qxidx < (int)pDetResult->pQx_ErrorList->size(); qxidx++)
+		{
+			QX_ERROR_INFO_ *QX_info = &pDetResult->pQx_ErrorList->at(qxidx);
+			// 只处理 盲失(Other) 的 NG 结果
+			if (QX_info->nconfig_qx_type != CONFIG_QX_NAME_Other || QX_info->result != QX_RESULT_TYPE_NG)
+			{
+				continue;
+			}
+
+			cv::Rect product_roi = QX_info->roi;
+			product_roi.x += fproduct_offx_piexl;
+			// 过滤通道同位置不存在盲失时，保留盲失NG
+			if (!IsOverlapQxInMsFilterChannel(pCheck_Result->ImageALLDetResultList,
+											  pImageResult->strChannel,
+											  product_roi,
+											  roi_expand, iou_threshold,
+											  pImageLog))
+			{
+				continue;
+			}
+
+			// 撤销NG结果
+			UndoNGResult(pImageResult, qxidx, QX_info);
+			nFilterNum++;
+
+			if (pImageLog)
+			{
+				pImageLog->AddCheckstr(PrintLevel_1, DET_LOG_LEVEL_3, "MsBlind",
+									   "img %s cam %s blob %d qx %s roi %s : 过滤通道同位置存在缺陷, 不算盲失",
+									   pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+									   qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str());
+			}
+			if (m_pQX_Merge_Analysis != nullptr && m_pQX_Merge_Analysis->m_pMergedetlog)
+			{
+				m_pQX_Merge_Analysis->m_pMergedetlog->AddCheckstr(PrintLevel_1, 3, "MsBlind",
+																  "img %s cam %s blob %d qx %s roi %s : 过滤通道同位置存在缺陷, 不算盲失",
+																  pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+																  qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str());
+			}
+		}
+	}
+
+	return nFilterNum;
+}
+
+/**
+ * @brief 跨通道同位置判断：判断当前缺陷(产品坐标系ROI)在过滤通道是否存在同位置缺陷
+ *        (match_qx_types 为空表示不限制缺陷类型)
+ *        同位置判断方式参考暗点联合分析：当前ROI与过滤通道缺陷ROI双方均按 expand 扩张后计算 IOU
+ * @param imgResults 当前相机所有通道的检测结果
+ * @param cur_channel_name 当前缺陷所属通道(主通道)
+ * @param filter_channels 需要比较的过滤通道
+ * @param match_qx_types 过滤通道中参与比较的缺陷类型(CONFIG_QX_NAME_)，为空表示不限制类型
+ * @param qx_product_roi 当前缺陷的产品坐标系ROI
+ * @param roi_expand 扩张像素
+ * @param iou_threshold IOU阈值
+ * @param strTag 日志标签
+ * @param plog 日志
+ * @return bool 过滤通道同位置存在匹配缺陷返回true
+ */
+bool ImageResultJudge::IsOverlapQxInFilterChannels(const std::vector<std::shared_ptr<ImageAllResult>> &imgResults,
+												   const std::string &cur_channel_name,
+												   const std::vector<std::string> &filter_channels,
+												   const std::vector<int> &match_qx_types,
+												   const cv::Rect &qx_product_roi,
+												   int roi_expand, float iou_threshold,
+												   const std::string &strTag,
+												   std::shared_ptr<DetLog> plog)
+{
+	// 防御：ROI非法/未配置过滤通道时不判断，避免无效计算与除零
+	if (qx_product_roi.width <= 0 || qx_product_roi.height <= 0)
+	{
+		return false;
+	}
+	// match_qx_types 为空表示不限制过滤通道的缺陷类型
+	if (filter_channels.empty())
+	{
+		return false;
+	}
+	if (m_pAnalysisyConfig == nullptr)
+	{
+		return false;
+	}
+	float fproduct_offx_mm = m_pAnalysisyConfig->commonCheckConfig.baseConfig.fProduct_Off_X_mm;
+	int fproduct_offx_piexl = int(fproduct_offx_mm / m_fImgage_Scale_X);
+
+	// 与暗点联合分析保持一致：双方均按 expand 扩张后再计算 IOU
+	cv::Rect roi_expanded(qx_product_roi.x - roi_expand, qx_product_roi.y - roi_expand,
+						  qx_product_roi.width + roi_expand * 2, qx_product_roi.height + roi_expand * 2);
+
+	for (int imgidx = 0; imgidx < (int)imgResults.size(); imgidx++)
+	{
+		std::shared_ptr<ImageAllResult> pFilterImage = imgResults.at(imgidx);
+		if (pFilterImage == nullptr)
+		{
+			continue;
+		}
+		// 只与配置的过滤通道比较；当前通道不参与，避免自身重叠导致误过滤
+		if (pFilterImage->strChannel == cur_channel_name)
+		{
+			continue;
+		}
+		if (std::find(filter_channels.begin(), filter_channels.end(), pFilterImage->strChannel) == filter_channels.end())
+		{
+			continue;
+		}
+		std::shared_ptr<One_Image_CheckResult_> pFilterDetResult = pFilterImage->pDetResult;
+		if (pFilterDetResult == nullptr || pFilterDetResult->pQx_ErrorList == nullptr)
+		{
+			continue;
+		}
+
+		// 只统计过滤通道内匹配类型的缺陷(未指定类型时不限制)，与具体判定结果无关
+		for (int j = 0; j < (int)pFilterDetResult->pQx_ErrorList->size(); j++)
+		{
+			const QX_ERROR_INFO_ &filter_qx = pFilterDetResult->pQx_ErrorList->at(j);
+			if (!match_qx_types.empty() &&
+				std::find(match_qx_types.begin(), match_qx_types.end(), filter_qx.nconfig_qx_type) == match_qx_types.end())
+			{
+				continue;
+			}
+			cv::Rect filter_roi = filter_qx.roi;
+			filter_roi.x += fproduct_offx_piexl;
+			cv::Rect filter_expanded(filter_roi.x - roi_expand, filter_roi.y - roi_expand,
+									 filter_roi.width + roi_expand * 2, filter_roi.height + roi_expand * 2);
+
+			float fiou = CheckUtil::CalIoU(roi_expanded, filter_expanded);
+			if (plog)
+			{
+				plog->AddCheckstr(PrintLevel_3, DET_LOG_LEVEL_3, strTag,
+								  "main channel %s roi %s <-> filter channel %s qx %d roi %s expand %d iou %f",
+								  cur_channel_name.c_str(), CheckUtil::GetRectString(qx_product_roi).c_str(),
+								  pFilterImage->strChannel.c_str(), j,
+								  CheckUtil::GetRectString(filter_roi).c_str(), roi_expand, fiou);
+			}
+			// IOU 为 NaN(ROI 非法导致的除零)时比较结果为 false，不会误过滤
+			if (fiou > iou_threshold)
+			{
+				if (plog)
+				{
+					plog->AddCheckstr(PrintLevel_1, DET_LOG_LEVEL_3, strTag,
+									  "main channel %s roi %s overlap filter channel %s qx %d roi %s expand %d iou %f > %f",
+									  cur_channel_name.c_str(), CheckUtil::GetRectString(qx_product_roi).c_str(),
+									  pFilterImage->strChannel.c_str(), j,
+									  CheckUtil::GetRectString(filter_roi).c_str(), roi_expand, fiou, iou_threshold);
+				}
+				return true;
+			}
+		}
+	}
+
+	return false;
+}
+
+/**
+ * @brief 跨通道联合判断(通用)：主通道中指定类型的NG缺陷，若在过滤通道同位置存在指定类型的缺陷，则撤销该NG
+ *        同位置判断方式参考暗点联合分析：双方按 expand 扩张后计算 IOU, 大于 iou_threshold 认为同位置
+ * @param pCheck_Result 当前相机检测结果(包含该相机所有通道的检测结果)
+ * @param main_channels 主通道
+ * @param filter_channels 过滤通道
+ * @param main_qx_types 参与判断的主通道缺陷类型(CONFIG_QX_NAME_)
+ * @param match_qx_types 过滤通道中参与比较的缺陷类型(CONFIG_QX_NAME_)，为空表示不限制类型
+ * @param roi_expand 扩张像素
+ * @param iou_threshold IOU阈值
+ * @param strTag 日志标签
+ * @param strReason 过滤原因(日志)
+ * @return int 被过滤的NG个数
+ */
+int ImageResultJudge::FilterCrossChannelNG(std::shared_ptr<CameraResult> pCheck_Result,
+										   const std::vector<std::string> &main_channels,
+										   const std::vector<std::string> &filter_channels,
+										   const std::vector<int> &main_qx_types,
+										   const std::vector<int> &match_qx_types,
+										   int roi_expand, float iou_threshold,
+										   const std::string &strTag,
+										   const std::string &strReason)
+{
+	if (pCheck_Result == nullptr || m_pAnalysisyConfig == nullptr)
+	{
+		return 0;
+	}
+	// 未配置主通道/过滤通道/判断类型时不需要判断
+	if (main_channels.empty() || filter_channels.empty() || main_qx_types.empty())
+	{
+		return 0;
+	}
+	float fproduct_offx_mm = m_pAnalysisyConfig->commonCheckConfig.baseConfig.fProduct_Off_X_mm;
+	int fproduct_offx_piexl = int(fproduct_offx_mm / m_fImgage_Scale_X);
+
+	int nFilterNum = 0;
+	for (int imgidx = 0; imgidx < (int)pCheck_Result->ImageALLDetResultList.size(); imgidx++)
+	{
+		std::shared_ptr<ImageAllResult> pImageResult = pCheck_Result->ImageALLDetResultList.at(imgidx);
+		if (pImageResult == nullptr)
+		{
+			continue;
+		}
+		// 只处理配置的主通道
+		if (std::find(main_channels.begin(), main_channels.end(), pImageResult->strChannel) == main_channels.end())
+		{
+			continue;
+		}
+		std::shared_ptr<One_Image_CheckResult_> pDetResult = pImageResult->pDetResult; // 检测结果
+		if (pDetResult == nullptr || pDetResult->pQx_ErrorList == nullptr || pImageResult->result == nullptr)
+		{
+			continue;
+		}
+		std::shared_ptr<DetLog> pImageLog = pImageResult->detlog;
+
+		for (int qxidx = 0; qxidx < (int)pDetResult->pQx_ErrorList->size(); qxidx++)
+		{
+			QX_ERROR_INFO_ *QX_info = &pDetResult->pQx_ErrorList->at(qxidx);
+			// 只处理指定类型的 NG 结果(YS 不降级)
+			if (QX_info->result != QX_RESULT_TYPE_NG)
+			{
+				continue;
+			}
+			if (std::find(main_qx_types.begin(), main_qx_types.end(), QX_info->nconfig_qx_type) == main_qx_types.end())
+			{
+				continue;
+			}
+
+			cv::Rect product_roi = QX_info->roi;
+			product_roi.x += fproduct_offx_piexl;
+			// 过滤通道同位置不存在匹配类型的缺陷时，保留该NG
+			if (!IsOverlapQxInFilterChannels(pCheck_Result->ImageALLDetResultList,
+											 pImageResult->strChannel,
+											 filter_channels, match_qx_types,
+											 product_roi,
+											 roi_expand, iou_threshold,
+											 strTag, pImageLog))
+			{
+				continue;
+			}
+
+			// 撤销NG结果
+			UndoNGResult(pImageResult, qxidx, QX_info);
+			nFilterNum++;
+
+			if (pImageLog)
+			{
+				pImageLog->AddCheckstr(PrintLevel_1, DET_LOG_LEVEL_3, strTag,
+									   "img %s cam %s blob %d qx %s roi %s : %s",
+									   pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+									   qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str(),
+									   strReason.c_str());
+			}
+			if (m_pQX_Merge_Analysis != nullptr && m_pQX_Merge_Analysis->m_pMergedetlog)
+			{
+				m_pQX_Merge_Analysis->m_pMergedetlog->AddCheckstr(PrintLevel_1, 3, strTag,
+																  "img %s cam %s blob %d qx %s roi %s : %s",
+																  pImageResult->strChannel.c_str(), pImageResult->cameraBaseResult->strCameraName.c_str(),
+																  qxidx, QX_info->qx_name.c_str(), CheckUtil::GetRectString(product_roi).c_str(),
+																  strReason.c_str());
+			}
+		}
+	}
+
+	return nFilterNum;
+}
+
+/**
+ * @brief cell黑白点联合判断：主通道(cellbw_main_channel)检出的cell黑白点(127画面的 Cell127 也算白点)
+ *        在过滤通道(cellbw_filter_channels)同位置存在缺陷(不限制缺陷类型)时不算cell黑白点(撤销NG)
+ *        同位置判断方式参考暗点联合分析：双方按 cellbw_expand 扩张后计算 IOU, 大于 cellbw_iou 认为同位置
+ * @param pCheck_Result 当前相机检测结果(包含该相机所有通道的检测结果)
+ * @return int 被过滤的cell黑白点NG个数
+ */
+int ImageResultJudge::FilterCellBwNG(std::shared_ptr<CameraResult> pCheck_Result)
+{
+	if (m_pbaseCheckFunction == nullptr)
+	{
+		return 0;
+	}
+	Base_Function_CellBW_Check *pCellBwParam = &m_pbaseCheckFunction->cellbw_check;
+	// 未开启cell黑白点分析时不需要判断(主通道/过滤通道为空时由通用逻辑直接返回)
+	if (!pCellBwParam->bOpen)
+	{
+		return 0;
+	}
+	// 主通道只处理 cell 黑白点(白点/黑点都算，Cell127 按白点处理)；过滤通道不限制缺陷类型，同位置存在任意类型缺陷即过滤
+	const std::vector<int> qxTypes = {CONFIG_QX_NAME_white_Cell, CONFIG_QX_NAME_black_Cell, CONFIG_QX_NAME_127Cell};
+	return FilterCrossChannelNG(pCheck_Result,
+								pCellBwParam->cellbw_main_channel, pCellBwParam->cellbw_filter_channels,
+								qxTypes, std::vector<int>(),
+								(int)(round)(pCellBwParam->CellBW_Expand), pCellBwParam->CellBW_IOU,
+								"CellBw", "过滤通道同位置存在缺陷, 不算cell黑白点");
+}
+
 int ImageResultJudge::MergeResult(std::shared_ptr<ImageAllResult> pImageResult, QX_Analysis_Result_List *ptemre)
 {
 
@@ -769,6 +1407,9 @@ int ImageResultJudge::MergeResult(std::shared_ptr<ImageAllResult> pImageResult,
 	}
 	int qxNum = pDetResult->pQx_ErrorList->size();
 
+	// 关联过滤源过滤：撤销与关联通道过滤源缺陷重叠的暗点NG(含单图判定的单点NG)
+	FilterCorrelatedDirtyNG(pImageResult);
+
 	for (int iqx = 0; iqx < (int)ptemre->resultList.size(); iqx++)
 	{
 		if (ptemre->resultList.at(iqx).channel_name != pImageResult->strChannel ||
@@ -814,7 +1455,8 @@ int ImageResultJudge::MergeResult(std::shared_ptr<ImageAllResult> pImageResult,
 			cv::Rect CutRoi = GetCutRoi(roi, pImageResult->detImg);
 			cv::Size sz = cv::Size(QX_SAMLLIMG_WIDTH, QX_SAMLLIMG_HEIGHT);
 
-			#ifdef AI_Time
+			if (NeedDrawAndSave())
+			{
 				if (pImageResult->detImg.channels() == 1)
 				{
 					cv::cvtColor(pImageResult->detImg(CutRoi), tem.srcImg, cv::COLOR_GRAY2BGR);
@@ -824,7 +1466,7 @@ int ImageResultJudge::MergeResult(std::shared_ptr<ImageAllResult> pImageResult,
 					tem.srcImg = pImageResult->detImg(CutRoi).clone();
 				}
 				cv::resize(tem.srcImg, tem.resizeImg, sz);
-			#endif
+			}
 
 			int nerrortype = 1;
 			int checkFlage = ANALYSIS_TYPE_TF;
@@ -885,7 +1527,10 @@ int ImageResultJudge::MergeResult(std::shared_ptr<ImageAllResult> pImageResult,
 					{
 						m_CheckResult_shareP->nresult = nqx_type;
 					}
+					if (NeedDrawAndSave())
+					{
 						GetAIDetImg(pImageResult, pCenter, tem.AI_in_Img, tem.AI_out_img);
+					}
 					m_CheckResult_shareP->qxImageResult.push_back(tem);
 				}
 			}
@@ -1257,14 +1902,14 @@ int ImageResultJudge::ConfigTypeToResultType(int nconfigType)
 {
 	bool bdesc = false;
 	int resultError_type = ERROR_TYPE_OK;
-	if(nconfigType < CONFIG_QX_NAME_Names.size())
+	if (nconfigType < CONFIG_QX_NAME_Names.size())
 	{
 		string cqn = CONFIG_QX_NAME_Names[nconfigType];
-		for(int i = 0; i < m_FlawCodeList.size(); i++)
+		for (int i = 0; i < m_FlawCodeList.size(); i++)
 		{
-			for(int j = 0; j < m_FlawCodeList.at(i).config_flaw_name.size(); j++)
+			for (int j = 0; j < m_FlawCodeList.at(i).config_flaw_name.size(); j++)
 			{
-				if(cqn == m_FlawCodeList.at(i).config_flaw_name[j])
+				if (cqn == m_FlawCodeList.at(i).config_flaw_name[j])
 				{
 					resultError_type = i;
 					bdesc = true;
@@ -1274,7 +1919,7 @@ int ImageResultJudge::ConfigTypeToResultType(int nconfigType)
 		}
 	}
 	// 如果描述没有配置，用默认值
-	if(!bdesc)
+	if (!bdesc)
 	{
 		switch (nconfigType)
 		{
@@ -1359,6 +2004,12 @@ int ImageResultJudge::ConfigTypeToResultType(int nconfigType)
 		case CONFIG_QX_NAME_LackPOL:
 			resultError_type = ERROR_TYPE_LackPol;
 			break;
+		case CONFIG_QX_NAME_RLine_X:
+			resultError_type = ERROR_TYPE_Line_X;
+			break;
+		case CONFIG_QX_NAME_RLine_Y:
+			resultError_type = ERROR_TYPE_Line_Y;
+			break;
 		default:
 			break;
 		}

ConfigModule/include/CheckConfigDefine.h

@@ -71,6 +71,8 @@ enum CONFIG_QX_NAME_
 	CONFIG_QX_NAME_black_Cell,		 // 白cell
 	CONFIG_QX_NAME_LD_HS,			 // 黑闪类画面的亮点
 	CONFIG_QX_NAME_LackPOL,			 // 缺失pol检测
+	CONFIG_QX_NAME_RLine_X,			 // 弱线 X
+	CONFIG_QX_NAME_RLine_Y,			 // 弱线 Y
 	CONFIG_QX_NAME_count,
 };
 // 缺陷项对应在参数中的名称
@@ -105,7 +107,9 @@ static std::vector<std::string> CONFIG_QX_NAME_Names =
 		"white_Cell",
 		"black_Cell",
 		"LD_HS",
-		"LackPOL"};
+		"LackPOL",
+		"RLine_X",
+		"RLine_Y"};
 
 // web控制检测的缺陷类型，
 enum Function_CONFIG_QX_NAME_
@@ -124,7 +128,7 @@ enum Function_CONFIG_QX_NAME_
 	Function_CONFIG_QX_NAME_PS,				  // ps(P0002)
 	Function_CONFIG_QX_NAME_Weak_Bright_Mura, // 白GAP(P1654)
 	Function_CONFIG_QX_NAME_No_Label,		  // 缺POL(P2833)
-	Function_CONFIG_QX_NAME_Other,			  // 缺POL(P2833)
+	Function_CONFIG_QX_NAME_Other,			  // 盲失
 	Function_CONFIG_QX_NAME_Chess,			  // Chess 异常
 	Function_CONFIG_QX_NAME_127Cell,		  // 白cell
 	Function_CONFIG_QX_NAME_Cell,			  // 独立检测的黑白cell点
@@ -205,8 +209,7 @@ static const std::string IMG_CHANNEL_NAME[] =
 		"RGB-HGRAY",
 		"SCANR",
 		"SCANG",
-		"SCANB"
-	};
+		"SCANB"};
 // 分析类型
 enum ANALYSIS_TYPE_
 {
@@ -466,6 +469,7 @@ struct BasicConfig
 		height_max = 999999;
 		bDrawShieldRoi = false;
 		bShield_ZF = false;
+		bShield_BQ = true;
 		bDrawPreRoi = false;
 		fUP_IOU = 0.9;
 		bCal_ImageScale = false;
@@ -794,6 +798,7 @@ struct Function_YXDet
 {
 	bool bOpen; // 是否开启
 	std::string strModle;
+	int minArea;
 	Function_YXDet()
 	{
 		Init();
@@ -802,20 +807,22 @@ struct Function_YXDet
 	{
 		bOpen = false;
 		strModle = "";
+		minArea = 0;
 	}
 	void copy(Function_YXDet tem)
 	{
 		this->bOpen = tem.bOpen;
 		this->strModle = tem.strModle;
+		this->minArea = tem.minArea;
 	}
 	void print(std::string str)
 	{
-		// printf("%s>>bOpen %d strModle %s\n", str.c_str(), bOpen, strModle.c_str());
+		// printf("%s>>bOpen %d strModle %s minArea %d\n", str.c_str(), bOpen, strModle.c_str(), minArea);
 	}
 	std::string GetInfo(std::string str)
 	{
 		char buffer[64];
-		sprintf(buffer, "%s>>bOpen %d strModle %s\n", str.c_str(), bOpen, strModle.c_str());
+		sprintf(buffer, "%s>>bOpen %d strModle %s minArea %d\n", str.c_str(), bOpen, strModle.c_str(), minArea);
 		std::string str123 = buffer;
 		return str123;
 	}
@@ -869,6 +876,12 @@ struct Function_AI_LD
 	bool bWTBLD;   // 是否是WTB亮点
 	bool bHSLD;	   // 黑闪类型的亮点
 	float fDP_IOU; // DP的IOU
+
+	bool bUseLD_Standard; // 是否使用亮点标准化参数
+	float fLD_Area;		  // 亮点面积
+	float fLD_En;		  // 亮点的能量
+	float fLD_HJ;		  // 亮点灰阶
+	float fLD_Len;		  // 亮点长度
 	Function_AI_LD()
 	{
 		Init();
@@ -880,6 +893,11 @@ struct Function_AI_LD
 		bWTBLD = false;
 		bHSLD = false;
 		fDP_IOU = 0.1;
+		fLD_Area = -1;
+		fLD_En = -1;
+		fLD_HJ = -1;
+		fLD_Len = -1;
+		bUseLD_Standard = false;
 	}
 	void copy(Function_AI_LD tem)
 	{
@@ -888,6 +906,11 @@ struct Function_AI_LD
 		this->bWTBLD = tem.bWTBLD;
 		this->bHSLD = tem.bHSLD;
 		this->fDP_IOU = tem.fDP_IOU;
+		this->fLD_Area = tem.fLD_Area;
+		this->fLD_En = tem.fLD_En;
+		this->fLD_HJ = tem.fLD_HJ;
+		this->fLD_Len = tem.fLD_Len;
+		this->bUseLD_Standard = tem.bUseLD_Standard;
 	}
 	void print(std::string str)
 	{
@@ -1271,7 +1294,19 @@ struct Function_Image_Align
 			{
 				printf("pointArry1 == 0 \n\n\n");
 			}
-			feature_Mask = tem_feature_Mask(feature_Roi).clone();
+			// 防御：feature_Roi 相对图像尺寸 img_W x img_H 可能越界，
+			// 越界会导致 cv::Mat(Mat, Rect) 断言失败并抛 cv::Exception，进而 abort。
+			cv::Rect roi = feature_Roi & cv::Rect(0, 0, img_W, img_H);
+			if (roi.width <= 0 || roi.height <= 0)
+			{
+				printf("[ToMaskImg] WARNING feature_Roi (%d,%d,%d,%d) out of range, img %dx%d\n",
+					   feature_Roi.x, feature_Roi.y, feature_Roi.width, feature_Roi.height, img_W, img_H);
+				feature_Mask = tem_feature_Mask.clone();
+			}
+			else
+			{
+				feature_Mask = tem_feature_Mask(roi).clone();
+			}
 		}
 	}
 	void print(std::string str)
@@ -1519,69 +1554,113 @@ struct AD_Analysisy_S
 		return str123;
 	}
 };
-// 暗点检测功能
-struct Function_AD_Check
+// 暗点同位置判断(IOU)
+struct AD_Analysisy_IOU
 {
-
-	bool bOpen;					   // 是否开启
-	AD_S_Standard S_standard_3s;   // 3s 标准
-	AD_S_Standard S_standard_2s;   // 2s 标准
-	AD_S_Standard S_standard_1s;   // 1s 标准
-	AD_Analysisy_Num analysis_num; // 数量分析
-	AD_Analysisy_Dis analysis_dis; // 距离分析
-	AD_Analysisy_S analysis_s;	   // s 标准分析
-
-	Function_AD_Check()
+	float AD_IOU;	 // IOU 阈值
+	float AD_Expand; // 扩张阈值
+	AD_Analysisy_IOU()
 	{
 		Init();
 	}
 	void Init()
 	{
-		bOpen = false;
-		S_standard_3s.Init();
-		S_standard_2s.Init();
-		S_standard_1s.Init();
-		analysis_num.Init();
-		analysis_dis.Init();
-		analysis_s.Init();
+		AD_IOU = 0.15f;
+		AD_Expand = 0.0f;
 	}
-
-	void copy(Function_AD_Check tem)
+	void copy(AD_Analysisy_IOU tem)
 	{
-		this->bOpen = tem.bOpen;
-		this->S_standard_3s.copy(tem.S_standard_3s);
-		this->S_standard_2s.copy(tem.S_standard_2s);
-		this->S_standard_1s.copy(tem.S_standard_1s);
-		this->analysis_num.copy(tem.analysis_num);
-		this->analysis_dis.copy(tem.analysis_dis);
-		this->analysis_s.copy(tem.analysis_s);
+		this->AD_IOU = tem.AD_IOU;
+		this->AD_Expand = tem.AD_Expand;
 	}
-	void print(std::string str)
+	std::string GetInfo(std::string str)
 	{
-		printf("%s>>bOpen %d %s %s %s %s %s %s\n", str.c_str(),
-			   bOpen, S_standard_3s.GetInfo("3S").c_str(),
-			   S_standard_2s.GetInfo("2S").c_str(),
-			   S_standard_1s.GetInfo("1S").c_str(),
-			   analysis_num.GetInfo("analysis_num").c_str(),
-			   analysis_dis.GetInfo("analysis_dis").c_str(),
-			   analysis_s.GetInfo("analysis_s").c_str());
+		char buffer[256];
+		sprintf(buffer, "%s:AD_IOU %f AD_Expand %f;", str.c_str(),
+				AD_IOU, AD_Expand);
+		std::string str123 = buffer;
+		return str123;
+	}
+};
+struct AD_Analysisy_Bind
+{
+	bool bUse;
+	std::vector<std::string> main_channel_2s;
+	std::vector<std::string> second_channel_2s;
+	std::vector<std::string> main_channel_3s;
+	std::vector<std::string> second_channel_3s;
+	std::vector<std::string> correlated_channels;
+	// 关联过滤源缺陷类型名称(CONFIG_QX_NAME_Names 中的名称, 如 "Dirty_L0"/"Dirty_L1"/"Dirty_L2"/"AD")
+	// 仅这些类型的缺陷会被登记为过滤源(与 correlated_channels 配合使用); 为空时默认三种脏污(兼容旧配置)
+	std::vector<std::string> correlated_types;
+
+	AD_Analysisy_Bind()
+	{
+		Init();
+	}
+	void Init()
+	{
+		bUse = false;
+		main_channel_2s.clear();
+		second_channel_2s.clear();
+		main_channel_3s.clear();
+		second_channel_3s.clear();
+		correlated_channels.clear();
+		correlated_types.clear();
+	}
+	void copy(AD_Analysisy_Bind tem)
+	{
+		this->bUse = tem.bUse;
+		this->main_channel_2s = tem.main_channel_2s;
+		this->second_channel_2s = tem.second_channel_2s;
+		this->main_channel_3s = tem.main_channel_3s;
+		this->second_channel_3s = tem.second_channel_3s;
+		this->correlated_channels = tem.correlated_channels;
+		this->correlated_types = tem.correlated_types;
 	}
 	std::string GetInfo(std::string str)
 	{
 		char buffer[256];
-		sprintf(buffer, "%s>>bOpen %d ", str.c_str(), bOpen);
+		sprintf(buffer, "%s:bUse %d;", str.c_str(),
+				bUse);
 		std::string str123 = buffer;
-		str123 += S_standard_3s.GetInfo("3S");
-		str123 += S_standard_2s.GetInfo("2S");
-		str123 += S_standard_1s.GetInfo("1S");
-		str123 += analysis_num.GetInfo("analysis_num");
-		str123 += analysis_dis.GetInfo("analysis_dis");
-		str123 += analysis_s.GetInfo("analysis_s");
-
 		return str123;
 	}
 };
 
+struct AD_Analysisy_SF
+{
+	std::vector<std::string> s_sf;
+	std::vector<std::string> region_sf;
+
+	AD_Analysisy_SF()
+	{
+		Init();
+	}
+	void Init()
+	{
+		s_sf.clear();
+		region_sf.clear();
+	}
+	void copy(AD_Analysisy_SF tem)
+	{
+		this->s_sf = tem.s_sf;
+		this->region_sf = tem.region_sf;
+	}
+	std::string GetInfo(std::string str)
+	{
+		std::string str123;
+		for (int i = 0; i < s_sf.size(); i++)
+		{
+			str123 += str + ":s_sf " + s_sf[i] + ";";
+		}
+		for (int i = 0; i < region_sf.size(); i++)
+		{
+			str123 += str + ":region_sf " + region_sf[i] + ";";
+		}
+		return str123;
+	}
+};
 // 异物检测功能
 struct Function_POL_Check
 {
@@ -1670,6 +1749,74 @@ struct Function_Edge_Det
 		return str123;
 	}
 };
+struct Function_Det_BQ
+{
+	bool bBQ_AI_Det; // 是否启用标签独立AI检测
+	int BQ_expand;
+	bool bsaveAll;
+
+	Function_Det_BQ()
+	{
+		Init();
+	}
+	void Init()
+	{
+		BQ_expand = 0;
+		bBQ_AI_Det = false;
+		bsaveAll = false;
+	}
+	void copy(Function_Det_BQ tem)
+	{
+		this->BQ_expand = tem.BQ_expand;
+		this->bBQ_AI_Det = tem.bBQ_AI_Det;
+		this->bsaveAll = tem.bsaveAll;
+	}
+	void print(std::string str)
+	{
+		printf("%d BQ_expand %d bBQ_AI_Det %d bsaveAll %d\n", str.c_str(),
+			   BQ_expand, bBQ_AI_Det, bsaveAll);
+	}
+	std::string GetInfo(std::string str)
+	{
+		char buffer[256];
+		sprintf(buffer, "%d BQ_expand %d bBQ_AI_Det %d bsaveAll %d\n", str.c_str(),
+				BQ_expand, bBQ_AI_Det, bsaveAll);
+		std::string str123 = buffer;
+		return str123;
+	}
+};
+
+// 启用127cell 独立检测
+struct Function_Det_RLine
+{
+	bool bOpen;		// 是否开启
+	bool bOpen_512; // 512模型是否开启。
+	Function_Det_RLine()
+	{
+		Init();
+	}
+	void Init()
+	{
+		bOpen = false;
+		bOpen_512 = false;
+	}
+	void copy(Function_Det_RLine tem)
+	{
+		this->bOpen = tem.bOpen;
+		this->bOpen_512 = tem.bOpen_512;
+	}
+	void print(std::string str)
+	{
+		printf("%s>>bOpen %d bOpen_512 %d \n", str.c_str(), bOpen, bOpen_512);
+	}
+	std::string GetInfo(std::string str)
+	{
+		char buffer[64];
+		sprintf(buffer, "%s>>bOpen %d bOpen_512 %d \n", str.c_str(), bOpen, bOpen_512);
+		std::string str123 = buffer;
+		return str123;
+	}
+};
 // 检测功能
 struct CheckFunction
 {
@@ -1686,9 +1833,10 @@ struct CheckFunction
 	Function_Image_Align f_Image_Align;		   // 图片特征对齐
 	Function_Detect_LackPol f_Dectect_LackPol; // 缺失POL检测
 	Function_SecondDet f_SecondDetect;		   // 二次检测分析
-	Function_AD_Check f_AD_Check;			   // 暗点的检测功能
 	Function_POL_Check f_POL_Check;			   // 异物的检测功能
 	Function_Edge_Det f_dege_Det;
+	Function_Det_BQ f_Det_BQ;
+	Function_Det_RLine f_RLineDet; // 弱线检测参数
 	CheckFunction()
 	{
 		Init();
@@ -1708,9 +1856,10 @@ struct CheckFunction
 		f_Image_Align.Init();
 		f_Dectect_LackPol.Init();
 		f_SecondDetect.Init();
-		f_AD_Check.Init();
 		f_POL_Check.Init();
 		f_dege_Det.Init();
+		f_Det_BQ.Init();
+		f_RLineDet.Init();
 	}
 	void copy(CheckFunction tem)
 	{
@@ -1728,9 +1877,10 @@ struct CheckFunction
 		this->f_Image_Align.copy(tem.f_Image_Align);
 		this->f_Dectect_LackPol.copy(tem.f_Dectect_LackPol);
 		this->f_SecondDetect.copy(tem.f_SecondDetect);
-		this->f_AD_Check.copy(tem.f_AD_Check);
 		this->f_POL_Check.copy(tem.f_POL_Check);
 		this->f_dege_Det.copy(tem.f_dege_Det);
+		this->f_Det_BQ.copy(tem.f_Det_BQ);
+		this->f_RLineDet.copy(tem.f_RLineDet);
 	}
 	void print(std::string str)
 	{
@@ -1748,9 +1898,10 @@ struct CheckFunction
 		f_Image_Align.print("Image_Align");
 		f_Dectect_LackPol.print("Dectect_LackPol");
 		f_SecondDetect.print("SecondDetect");
-		f_AD_Check.print("f_AD_Check");
 		f_POL_Check.print("f_POL_Check");
 		f_dege_Det.print("f_dege_Det");
+		f_Det_BQ.print("f_Det_BQ");
+		f_RLineDet.print("f_RLineDet");
 	}
 	std::string GetInfo(std::string str)
 	{
@@ -1769,9 +1920,10 @@ struct CheckFunction
 		str123 += f_Image_Align.GetInfo("Image_Align");
 		str123 += f_Dectect_LackPol.GetInfo("Dectect_LackPol");
 		str123 += f_SecondDetect.GetInfo("SecondDetect");
-		str123 += f_AD_Check.GetInfo("f_AD_Check");
 		str123 += f_POL_Check.GetInfo("f_POL_Check");
 		str123 += f_dege_Det.GetInfo("f_dege_Det");
+		str123 += f_Det_BQ.GetInfo("f_Det_BQ");
+		str123 += f_RLineDet.GetInfo("f_RLineDet");
 		return str123;
 	}
 };
@@ -2155,6 +2307,9 @@ struct Base_Function_AD_Check
 	AD_Analysisy_Num analysis_num;	 // 数量分析
 	AD_Analysisy_Dis analysis_dis;	 // 距离分析
 	AD_Analysisy_S analysis_s;		 // s 标准分析
+	AD_Analysisy_IOU analysis_iou;	 // IOU 分析
+	AD_Analysisy_Bind analysis_bind; // 绑定分析
+	AD_Analysisy_SF analysis_sf;	 // 同画面参数分析
 
 	Base_Function_AD_Check()
 	{
@@ -2169,6 +2324,9 @@ struct Base_Function_AD_Check
 		analysis_num.Init();
 		analysis_dis.Init();
 		analysis_s.Init();
+		analysis_iou.Init();
+		analysis_bind.Init();
+		analysis_sf.Init();
 	}
 
 	void copy(Base_Function_AD_Check tem)
@@ -2180,16 +2338,22 @@ struct Base_Function_AD_Check
 		this->analysis_num.copy(tem.analysis_num);
 		this->analysis_dis.copy(tem.analysis_dis);
 		this->analysis_s.copy(tem.analysis_s);
+		this->analysis_iou.copy(tem.analysis_iou);
+		this->analysis_bind.copy(tem.analysis_bind);
+		this->analysis_sf.copy(tem.analysis_sf);
 	}
 	void print(std::string str)
 	{
-		printf("%s>>bOpen %d %s %s %s %s %s %s\n", str.c_str(),
+		printf("%s>>bOpen %d %s %s %s %s %s %s %s %s %s\n", str.c_str(),
 			   bOpen, S_standard_3s.GetInfo("3S").c_str(),
 			   S_standard_2s.GetInfo("2S").c_str(),
 			   S_standard_1s.GetInfo("1S").c_str(),
 			   analysis_num.GetInfo("analysis_num").c_str(),
 			   analysis_dis.GetInfo("analysis_dis").c_str(),
-			   analysis_s.GetInfo("analysis_s").c_str());
+			   analysis_s.GetInfo("analysis_s").c_str(),
+			   analysis_iou.GetInfo("analysis_iou").c_str(),
+			   analysis_bind.GetInfo("analysis_bind").c_str(),
+			   analysis_sf.GetInfo("analysis_sf").c_str());
 	}
 	std::string GetInfo(std::string str)
 	{
@@ -2202,7 +2366,114 @@ struct Base_Function_AD_Check
 		str123 += analysis_num.GetInfo("analysis_num");
 		str123 += analysis_dis.GetInfo("analysis_dis");
 		str123 += analysis_s.GetInfo("analysis_s");
+		str123 += analysis_iou.GetInfo("analysis_iou");
+		str123 += analysis_bind.GetInfo("analysis_time");
+		str123 += analysis_sf.GetInfo("analysis_sf");
+
+		return str123;
+	}
+};
 
+// 盲失检测功能
+struct Base_Function_MS_Check
+{
+
+	bool bOpen;									  // 是否开启
+	float MS_IOU;								  // IOU 阈值(同位置判断)
+	float MS_Expand;							  // 扩张值(像素)
+	std::vector<std::string> ms_main_channel;	  // 主通道绑定
+	std::vector<std::string> ms_filter_channels;  // 过滤通道绑定
+
+	Base_Function_MS_Check()
+	{
+		Init();
+	}
+	void Init()
+	{
+		bOpen = false;
+		MS_IOU = 0.8f;
+		MS_Expand = 0.0f;
+		ms_main_channel.clear();
+		ms_filter_channels.clear();
+	}
+
+	void copy(Base_Function_MS_Check tem)
+	{
+		this->bOpen = tem.bOpen;
+		this->MS_IOU = tem.MS_IOU;
+		this->MS_Expand = tem.MS_Expand;
+		this->ms_main_channel = tem.ms_main_channel;
+		this->ms_filter_channels = tem.ms_filter_channels;
+	}
+	void print(std::string str)
+	{
+		printf("%s", GetInfo(str).c_str());
+	}
+	std::string GetInfo(std::string str)
+	{
+		char buffer[256];
+		sprintf(buffer, "%s>>bOpen %d MS_IOU %f MS_Expand %f ", str.c_str(), bOpen, MS_IOU, MS_Expand);
+		std::string str123 = buffer;
+		for (int i = 0; i < ms_main_channel.size(); i++)
+		{
+			str123 += str + ":ms_main_channel " + ms_main_channel.at(i) + ";";
+		}
+		for (int i = 0; i < ms_filter_channels.size(); i++)
+		{
+			str123 += str + ":ms_filter_channels " + ms_filter_channels.at(i) + ";";
+		}
+		str123 += "\n";
+		return str123;
+	}
+};
+
+// Cell黑白点检测功能
+struct Base_Function_CellBW_Check
+{
+	bool bOpen;										  // 是否开启
+	float CellBW_IOU;								  // IOU 阈值(同位置判断)
+	float CellBW_Expand;							  // 扩张值(像素)
+	std::vector<std::string> cellbw_main_channel;	  // 主通道绑定
+	std::vector<std::string> cellbw_filter_channels;  // 过滤通道绑定
+
+	Base_Function_CellBW_Check()
+	{
+		Init();
+	}
+	void Init()
+	{
+		bOpen = false;
+		CellBW_IOU = 0.8f;
+		CellBW_Expand = 0.0f;
+		cellbw_main_channel.clear();
+		cellbw_filter_channels.clear();
+	}
+	void copy(Base_Function_CellBW_Check tem)
+	{
+		this->bOpen = tem.bOpen;
+		this->CellBW_IOU = tem.CellBW_IOU;
+		this->CellBW_Expand = tem.CellBW_Expand;
+		this->cellbw_main_channel = tem.cellbw_main_channel;
+		this->cellbw_filter_channels = tem.cellbw_filter_channels;
+	}
+	void print(std::string str)
+	{
+		printf("%s", GetInfo(str).c_str());
+	}
+	std::string GetInfo(std::string str)
+	{
+		char buffer[256];
+		sprintf(buffer, "%s>>bOpen %d CellBW_IOU %f CellBW_Expand %f ", str.c_str(), bOpen, CellBW_IOU, CellBW_Expand);
+		std::string str123 = buffer;
+		for (int i = 0; i < cellbw_main_channel.size(); i++)
+		{
+			str123 += str + ":cellbw_main_channel " + cellbw_main_channel.at(i) + ";";
+		}
+		for (int i = 0; i < cellbw_filter_channels.size(); i++)
+		{
+			str123 += str + ":cellbw_filter_channels " + cellbw_filter_channels.at(i) + ";";
+		}
+		str123 += "\n";
 		return str123;
 	}
 };
@@ -2216,6 +2487,9 @@ struct BaseCheckFunction
 	Base_Function_BigNG bigNG;
 	Base_Function_EdgeChannel edgeChannel;
 	Base_Function_AD_Check ad_check;
+	Base_Function_MS_Check ms_check;
+	Base_Function_CellBW_Check cellbw_check;
+	bool bDrawAndSave; // 是否存图绘制（绘制并保存结果图，运行时替代 SAVE_RESIMG 宏）
 	BaseCheckFunction()
 	{
 		Init();
@@ -2228,6 +2502,9 @@ struct BaseCheckFunction
 		bigNG.Init();
 		edgeChannel.Init();
 		ad_check.Init();
+		ms_check.Init();
+		cellbw_check.Init();
+		bDrawAndSave = false;
 	}
 	void copy(BaseCheckFunction tem)
 	{
@@ -2237,6 +2514,9 @@ struct BaseCheckFunction
 		this->bigNG.copy(tem.bigNG);
 		this->edgeChannel.copy(tem.edgeChannel);
 		this->ad_check.copy(tem.ad_check);
+		this->ms_check.copy(tem.ms_check);
+		this->cellbw_check.copy(tem.cellbw_check);
+		this->bDrawAndSave = tem.bDrawAndSave;
 	}
 	void print(std::string str)
 	{
@@ -2247,6 +2527,9 @@ struct BaseCheckFunction
 		bigNG.print("bigNG");
 		edgeChannel.print("edgeChannel");
 		ad_check.print("ad_check");
+		ms_check.print("ms_check");
+		cellbw_check.print("cellbw_check");
+		printf("bDrawAndSave %d\n", bDrawAndSave);
 	}
 	std::string GetInfo(std::string str)
 	{
@@ -2257,6 +2540,8 @@ struct BaseCheckFunction
 		str123 += bigNG.GetInfo("bigNG");
 		str123 += edgeChannel.GetInfo("edgeChannel");
 		str123 += ad_check.GetInfo("ad_check");
+		str123 += ms_check.GetInfo("ms_check");
+		str123 += cellbw_check.GetInfo("cellbw_check");
 		// str123 += "\n";
 		return str123;
 	}

ConfigModule/include/JsonConfig.h

@@ -64,6 +64,8 @@ public:
     int GetConfig(ALLChannelCheckFunction &config);
     int GetFunction(Json::Value value, CheckFunction &function);
 
+    static bool s_bDetTag; // 标签检测模式标志，由命令行 -tag 参数控制
+
 private:
     ALLChannelCheckFunction _config;
 };

ConfigModule/src/ConfigInstance.cpp

@@ -213,6 +213,7 @@ int ConfigInstance::Updata_Check(Json::Value json_value)
 
 int ConfigInstance::SetStatus(int nConfigType)
 {
+    std::lock_guard<std::mutex> lock(mutex_status);
     for (int i = 0; i < MAX_USER_COUNT; i++)
     {
         m_USER_ConfigUpdataStatusList[nConfigType][i] = true;

ConfigModule/src/ConfigManager.cpp

@@ -22,62 +22,6 @@ ConfigManager::~ConfigManager()
 {
 }
 
-std::vector <std::string> QX_Result_Names =
-	{
-		"OK",
-		"AD_YX",
-		"X_Line",
-		"Y_Line",
-		"fangge",
-		"Rubbing_Mura",
-		"Broken_line",
-		"ZARA",
-		"MTX",
-		"POL_Cell",
-		"Bright_Point",
-		"Dark_Point",
-		"BLack_Point",
-		"White_Point",
-		"Scratch",
-		"Weak_Bright_Mura",
-		"No_Label",
-		"Bright_Mura_Exe",
-		"Sweak_Line_Dark",
-		"STEAM_POCKET",
-		"Dirty",
-		"other",
-		"Cell_W",
-		"Cell_B",
-		"LackPol"};
-std::vector <std::string> QX_Result_Code =
-	{
-		"P1153",
-		"P6873",
-		"P3351",
-		"P3452",
-		"P3453",
-		"P1550",
-		"P3379",
-		"P1153",
-		"P1164",
-		"P1101",
-		"P1112",
-		"P1111",
-		"P1104",
-		"P1103",
-		"P1557",
-		"P1654",
-		"P2833",
-		"P1549",
-		"P1204",
-		"P2534",
-		"P2534",
-		"P1101",
-		"P1103",
-		"P1104",
-		"P8001",
-};
-
 int ReadFlawCodeConfig(std::string json_path)
 {
 	m_FlawCodeList.erase(m_FlawCodeList.begin(), m_FlawCodeList.end());
@@ -203,6 +147,7 @@ int ConfigManager::UpdateConfig()
 		if (!std::filesystem::exists(strConfigPath)) {
     		strConfigPath = m_strConfigRootPath + "/param.json";
 		}
+		printf("read param %s\n", strConfigPath.c_str());
 		std::shared_ptr<ConfigBase> temConfig = ConfigBase::GetInstance();
 		re = LoadParamConfig(temConfig, strConfigPath);
 		if (re != 0)
@@ -218,8 +163,13 @@ int ConfigManager::UpdateConfig()
 			printf("Error >>>> camear Name is empty \n");
 			continue;
 		}
+		auto it = Config_instances_.find(p.commonCheckConfig.baseConfig.strCamearName);
+		if (it != Config_instances_.end() && it->second) {
+		LoadParamConfig(it->second, strConfigPath);
+		} else {
 			Config_instances_[p.commonCheckConfig.baseConfig.strCamearName] = temConfig;
 		}
+	}
 
 	// getchar();
 

ConfigModule/src/JsonConfig.cpp

@@ -1,5 +1,8 @@
 #include "JsonConfig.h"
 
+// 静态成员定义：标签检测模式标志
+bool ChannelFuntonConfigJson::s_bDetTag = false;
+
 Json::Value CommonParamJson::toJsonValue()
 {
     return Json::Value();
@@ -32,7 +35,7 @@ void CommonParamToCheckConfigJson::toObjectFromValue(Json::Value root)
     std::unique_ptr<Json::CharReader> reader(builder.newCharReader());
     Json::Value rootvalue;
     std::string err;
-    // std::cout << strJson << std::endl;
+    std::cout << strJson << std::endl;
     auto nSize = strJson.size();
     if (reader->parse(strJson.c_str(), strJson.c_str() + nSize, &rootvalue, &err))
     {
@@ -45,10 +48,11 @@ void CommonParamToCheckConfigJson::toObjectFromValue(Json::Value root)
             if (value.isObject())
             {
                 _config.baseConfig.strConfigVersion = value["version"].asString();
-                if ( _config.baseConfig.strConfigVersion == "")
+                if (_config.baseConfig.strConfigVersion == "")
                 {
                     _config.baseConfig.strConfigVersion = "NULL";
                 }
+
                 _config.baseConfig.image_widht = value["image_widht"].asInt();
                 _config.baseConfig.Image_height = value["Image_height"].asInt();
                 _config.baseConfig.bDrawShieldRoi = value["bDrawShieldRoi"].asBool();
@@ -530,6 +534,10 @@ int ChannelFuntonConfigJson::GetFunction(Json::Value value, CheckFunction &funct
                 {
                     function.f_YXDet.strModle = value_f["form"]["YX_Det"]["AI_Model"].asString();
                 }
+                if (value_f["form"]["YX_Det"]["minArea"])
+                {
+                    function.f_YXDet.minArea = value_f["form"]["YX_Det"]["minArea"].asInt();
+                }
 
                 // function.f_UseUpQX.print("UpQX");
             }
@@ -559,6 +567,27 @@ int ChannelFuntonConfigJson::GetFunction(Json::Value value, CheckFunction &funct
                     {
                         function.f_LDConfig.fDP_IOU = value_f["form"]["LD_Det"]["DP_IOU"].asFloat();
                     }
+
+                    if (value_f["form"]["LD_Standard"]["LD_Area"])
+                    {
+                        function.f_LDConfig.fLD_Area = value_f["form"]["LD_Standard"]["LD_Area"].asFloat();
+                    }
+                    if (value_f["form"]["LD_Standard"]["LD_En"])
+                    {
+                        function.f_LDConfig.fLD_En = value_f["form"]["LD_Standard"]["LD_En"].asFloat();
+                    }
+                    if (value_f["form"]["LD_Standard"]["LD_HJ"])
+                    {
+                        function.f_LDConfig.fLD_HJ = value_f["form"]["LD_Standard"]["LD_HJ"].asFloat();
+                    }
+                    if (value_f["form"]["LD_Standard"]["LD_Len"])
+                    {
+                        function.f_LDConfig.fLD_Len = value_f["form"]["LD_Standard"]["LD_Len"].asFloat();
+                    }
+                    if (value_f["form"]["LD_Standard"]["bUseLD_Standard"])
+                    {
+                        function.f_LDConfig.bUseLD_Standard = value_f["form"]["LD_Standard"]["bUseLD_Standard"].asBool();
+                    }
                 }
                 else
                 {
@@ -953,83 +982,6 @@ int ChannelFuntonConfigJson::GetFunction(Json::Value value, CheckFunction &funct
                 }
                 // getchar();
             }
-            // 暗点检测
-            if ("AD_Check" == strCode)
-            {
-                auto value_f = value[i];
-                // std::cout << value_f << std::endl;
-                function.f_AD_Check.bOpen = value_f["isOpen"].asBool();
-                if (function.f_AD_Check.bOpen)
-                {
-                    if (value_f["form"]["AD_S_Standard"]["AD_3S_Area"])
-                    {
-                        function.f_AD_Check.S_standard_3s.area = value_f["form"]["AD_S_Standard"]["AD_3S_Area"].asFloat();
-                    }
-                    if (value_f["form"]["AD_S_Standard"]["AD_3S_Len"])
-                    {
-                        function.f_AD_Check.S_standard_3s.len = value_f["form"]["AD_S_Standard"]["AD_3S_Len"].asFloat();
-                    }
-
-                    if (value_f["form"]["AD_S_Standard"]["AD_2S_Area"])
-                    {
-                        function.f_AD_Check.S_standard_2s.area = value_f["form"]["AD_S_Standard"]["AD_2S_Area"].asFloat();
-                    }
-                    if (value_f["form"]["AD_S_Standard"]["AD_2S_Len"])
-                    {
-                        function.f_AD_Check.S_standard_2s.len = value_f["form"]["AD_S_Standard"]["AD_2S_Len"].asFloat();
-                    }
-
-                    if (value_f["form"]["AD_S_Standard"]["AD_1S_Area"])
-                    {
-                        function.f_AD_Check.S_standard_1s.area = value_f["form"]["AD_S_Standard"]["AD_1S_Area"].asFloat();
-                    }
-                    if (value_f["form"]["AD_S_Standard"]["AD_1S_Len"])
-                    {
-                        function.f_AD_Check.S_standard_1s.len = value_f["form"]["AD_S_Standard"]["AD_1S_Len"].asFloat();
-                    }
-
-                    if (value_f["form"]["AD_Check_Num"]["Open"])
-                    {
-                        function.f_AD_Check.analysis_num.bOpen = value_f["form"]["AD_Check_Num"]["Open"].asBool();
-                    }
-                    if (value_f["form"]["AD_Check_Num"]["Num"])
-                    {
-                        function.f_AD_Check.analysis_num.numT = value_f["form"]["AD_Check_Num"]["Num"].asInt();
-                    }
-
-                    if (value_f["form"]["AD_Check_Dis"]["Open"])
-                    {
-                        function.f_AD_Check.analysis_dis.bOpen = value_f["form"]["AD_Check_Dis"]["Open"].asBool();
-                    }
-                    if (value_f["form"]["AD_Check_Dis"]["Dis"])
-                    {
-                        function.f_AD_Check.analysis_dis.disT = value_f["form"]["AD_Check_Dis"]["Dis"].asFloat();
-                    }
-
-                    if (value_f["form"]["AD_Check_S"]["Open"])
-                    {
-                        function.f_AD_Check.analysis_s.bOpen = value_f["form"]["AD_Check_S"]["Open"].asBool();
-                    }
-
-                    if (value_f["form"]["AD_Check_S"]["S_value"])
-                    {
-                        function.f_AD_Check.analysis_s.Check_s_Value = value_f["form"]["AD_Check_S"]["S_value"].asInt();
-                    }
-                    if (value_f["form"]["AD_Check_S"]["NG_3s"])
-                    {
-                        function.f_AD_Check.analysis_s.NG_3s = value_f["form"]["AD_Check_S"]["NG_3s"].asBool();
-                    }
-                    if (value_f["form"]["AD_Check_S"]["Num"])
-                    {
-                        function.f_AD_Check.analysis_s.Check_s_Num = value_f["form"]["AD_Check_S"]["Num"].asInt();
-                    }
-                }
-                else
-                {
-                    function.f_AD_Check.Init();
-                }
-                // getchar();
-            }
             // 异物检测
             if ("POL_Cam" == strCode)
             {
@@ -1083,6 +1035,50 @@ int ChannelFuntonConfigJson::GetFunction(Json::Value value, CheckFunction &funct
                 // function.f_markDet.print("UpQX");
                 // getchar();
             }
+            // 标签配置参数
+            if ("Det_BQ" == strCode)
+            {
+                auto value_f = value[i];
+                // std::cout << value_f << std::endl;
+                // getchar();
+                function.f_Det_BQ.bBQ_AI_Det = value_f["isOpen"].asBool();
+                {
+                    if (value_f["form"]["BQ_Param"]["BQ_expand"])
+                    {
+                        function.f_Det_BQ.BQ_expand = value_f["form"]["BQ_Param"]["BQ_expand"].asInt();
+                    }
+                    if (value_f["form"]["BQ_Param"]["bsaveAll"])
+                    {
+                        function.f_Det_BQ.bsaveAll = value_f["form"]["BQ_Param"]["bsaveAll"].asBool();
+                    }
+                }
+
+                // 当 -tag 参数启用时，强制开启标签 AI 检测和保存所有中间结果
+                if (s_bDetTag)
+                {
+                    function.f_Det_BQ.bBQ_AI_Det = true;
+                    function.f_Det_BQ.bsaveAll = true;
+                }
+            }
+            if ("Det_RLine" == strCode)
+            {
+                auto value_f = value[i];
+                // std::cout << value_f << std::endl;
+                function.f_RLineDet.bOpen = value_f["isOpen"].asBool();
+                if (function.f_RLineDet.bOpen)
+                {
+                    if (value_f["form"]["Config_Model_512"]["bOpen"])
+                    {
+                        function.f_RLineDet.bOpen_512 = value_f["form"]["Config_Model_512"]["bOpen"].asBool();
+                    }
+                }
+                else
+                {
+                    function.f_RLineDet.Init();
+                }
+                // function.f_RLineDet.print("");
+                // getchar();
+            }
         }
     }
     return 0;
@@ -1375,6 +1371,7 @@ int BaseFuntonConfigJson::GetFunction(Json::Value value)
         _config.ad_check.bOpen = value_f["isOpen"].asBool();
         if (_config.ad_check.bOpen)
         {
+            // S标准
             if (value_f["form"]["AD_S_Standard"]["AD_3S_Area"])
             {
                 _config.ad_check.S_standard_3s.area = value_f["form"]["AD_S_Standard"]["AD_3S_Area"].asFloat();
@@ -1402,6 +1399,7 @@ int BaseFuntonConfigJson::GetFunction(Json::Value value)
                 _config.ad_check.S_standard_1s.len = value_f["form"]["AD_S_Standard"]["AD_1S_Len"].asFloat();
             }
 
+            // 数量分析
             if (value_f["form"]["AD_Check_Num"]["Open"])
             {
                 _config.ad_check.analysis_num.bOpen = value_f["form"]["AD_Check_Num"]["Open"].asBool();
@@ -1411,6 +1409,7 @@ int BaseFuntonConfigJson::GetFunction(Json::Value value)
                 _config.ad_check.analysis_num.numT = value_f["form"]["AD_Check_Num"]["Num"].asInt();
             }
 
+            // S数量分析
             if (value_f["form"]["AD_Check_Dis"]["Open"])
             {
                 _config.ad_check.analysis_dis.bOpen = value_f["form"]["AD_Check_Dis"]["Open"].asBool();
@@ -1420,6 +1419,7 @@ int BaseFuntonConfigJson::GetFunction(Json::Value value)
                 _config.ad_check.analysis_dis.disT = value_f["form"]["AD_Check_Dis"]["Dis"].asFloat();
             }
 
+            // 距离分析
             if (value_f["form"]["AD_Check_S"]["Open"])
             {
                 _config.ad_check.analysis_s.bOpen = value_f["form"]["AD_Check_S"]["Open"].asBool();
@@ -1437,12 +1437,196 @@ int BaseFuntonConfigJson::GetFunction(Json::Value value)
             {
                 _config.ad_check.analysis_s.Check_s_Num = value_f["form"]["AD_Check_S"]["Num"].asInt();
             }
+
+            // IOU分析
+            if (value_f["form"]["AD_Check_IOU"]["AD_IOU"])
+            {
+                _config.ad_check.analysis_iou.AD_IOU = value_f["form"]["AD_Check_IOU"]["AD_IOU"].asFloat();
+            }
+            if (value_f["form"]["AD_Check_IOU"]["AD_Expand"])
+            {
+                _config.ad_check.analysis_iou.AD_Expand = value_f["form"]["AD_Check_IOU"]["AD_Expand"].asFloat();
+            }
+
+            // 通道绑定
+            if (value_f["form"]["Bind_Channel"]["bUse_bind"])
+            {
+                _config.ad_check.analysis_bind.bUse = value_f["form"]["Bind_Channel"]["bUse_bind"].asBool();
+            }
+            auto value_qx = value_f["form"]["Bind_Channel"]["2s_main_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.main_channel_2s.push_back(channel);
+                }
+            }
+            value_qx = value_f["form"]["Bind_Channel"]["2s_second_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.second_channel_2s.push_back(channel);
+                }
+            }
+
+            value_qx = value_f["form"]["Bind_Channel"]["3s_main_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.main_channel_3s.push_back(channel);
+                }
+            }
+            value_qx = value_f["form"]["Bind_Channel"]["3s_second_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.second_channel_3s.push_back(channel);
+                }
+            }
+
+            value_qx = value_f["form"]["Bind_Channel"]["correlated_channels"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.correlated_channels.push_back(channel);
+                }
+            }
+
+            // 关联过滤源缺陷类型(如 ["Dirty_L0","Dirty_L1","Dirty_L2","AD"])
+            value_qx = value_f["form"]["Bind_Channel"]["correlated_types"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string qx_type = value_qx[idx].asString();
+                    _config.ad_check.analysis_bind.correlated_types.push_back(qx_type);
+                }
+            }
+
+            // 同画面分析
+            value_qx = value_f["form"]["Single_Channel"]["s_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_sf.s_sf.push_back(channel);
+                }
+            }
+            value_qx = value_f["form"]["Single_Channel"]["region1_channel"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ad_check.analysis_sf.region_sf.push_back(channel);
+                }
+            }
         }
         else
         {
             _config.ad_check.Init();
         }
     }
+    // 是否存图绘制开关：控制结果图绘制与保存（运行时替代 SAVE_RESIMG 宏）
+    if ("bDrawAndSave" == strCode)
+    {
+        _config.bDrawAndSave = value["isOpen"].asBool();
+    }
+    // 盲失分析(code绑定的qx_Other)
+    if ("MS_Analysis" == strCode)
+    {
+        auto value_f = value;
+        // std::cout << value_f << std::endl;
+        _config.ms_check.bOpen = value_f["isOpen"].asBool();
+        if (_config.ms_check.bOpen)
+        { 
+            // IOU分析
+            if (value_f["form"]["MS_Check_Param"]["ms_iou"])
+            {
+                _config.ms_check.MS_IOU = value_f["form"]["MS_Check_Param"]["ms_iou"].asFloat();
+            }
+            if (value_f["form"]["MS_Check_Param"]["ms_expand"])
+            {
+                _config.ms_check.MS_Expand = value_f["form"]["MS_Check_Param"]["ms_expand"].asFloat();
+            }
 
+            // 通道绑定
+            auto value_qx = value_f["form"]["MS_Check_Param"]["ms_main_channels"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ms_check.ms_main_channel.push_back(channel);
+                }
+            }
+            value_qx = value_f["form"]["MS_Check_Param"]["ms_filter_channels"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.ms_check.ms_filter_channels.push_back(channel);
+                }
+            }
+        }
+        else
+        {
+            _config.ms_check.Init();
+        }
+    }
+    // Cell黑白点分析(code绑定的qx_Other)
+    if ("Cellbw_Analysis" == strCode)
+    {
+        auto value_f = value;
+        // std::cout << value_f << std::endl;
+        _config.cellbw_check.bOpen = value_f["isOpen"].asBool();
+        if (_config.cellbw_check.bOpen)
+        { 
+            // IOU分析
+            if (value_f["form"]["CellBW_Check_Param"]["cellbw_iou"])
+            {
+                _config.cellbw_check.CellBW_IOU = value_f["form"]["CellBW_Check_Param"]["cellbw_iou"].asFloat();
+            }
+            if (value_f["form"]["CellBW_Check_Param"]["cellbw_expand"])
+            {
+                _config.cellbw_check.CellBW_Expand = value_f["form"]["CellBW_Check_Param"]["cellbw_expand"].asFloat();
+            }
+
+            // 通道绑定
+            auto value_qx = value_f["form"]["CellBW_Check_Param"]["cellbw_main_channels"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.cellbw_check.cellbw_main_channel.push_back(channel);
+                }
+            }
+            value_qx = value_f["form"]["CellBW_Check_Param"]["cellbw_filter_channels"];
+            if (value_qx.isArray())
+            {
+                for (int idx = 0; idx < value_qx.size(); idx++)
+                {
+                    std::string channel = value_qx[idx].asString();
+                    _config.cellbw_check.cellbw_filter_channels.push_back(channel);
+                }
+            }
+        }
+        else
+        {
+            _config.cellbw_check.Init();
+        }
+    }
     return 0;
 }

example/CMakeLists.txt

@@ -14,6 +14,7 @@ include_directories(
 ${CMAKE_CURRENT_SOURCE_DIR}/include
 ${PROJECT_SOURCE_DIR}/AlgorithmModule/include
 ${PROJECT_SOURCE_DIR}/ConfigModule/include
+${PROJECT_SOURCE_DIR}/AIEngineModule/include_base
 )
 
 link_directories(

example/MemMonitor.h

@@ -0,0 +1,177 @@
+#ifndef _MEM_MONITOR_H_
+#define _MEM_MONITOR_H_
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/time.h>
+#include <sys/resource.h>
+
+// 获取当前进程 RSS (物理内存) 使用量，单位 MB
+static inline long getRSS_KB()
+{
+    long rss = 0;
+    FILE *fp = fopen("/proc/self/status", "r");
+    if (!fp) return -1;
+    
+    char line[256];
+    while (fgets(line, sizeof(line), fp))
+    {
+        if (strncmp(line, "VmRSS:", 6) == 0)
+        {
+            // 格式: "VmRSS:    12345 kB"
+            const char *p = line + 6;
+            while (*p == ' ' || *p == '\t') p++;
+            rss = atol(p);
+            break;
+        }
+    }
+    fclose(fp);
+    return rss;
+}
+
+// 获取当前进程内存使用，打印格式化日志
+static inline void printMemUsage(const char *tag, const char *extra)
+{
+    long rss = getRSS_KB();
+    struct rusage usage;
+    getrusage(RUSAGE_SELF, &usage);
+    printf("[MEM] %-30s | VmRSS: %6ld MB | maxRSS: %6ld MB | %s\n",
+           tag, rss / 1024, usage.ru_maxrss / 1024, extra ? extra : "");
+}
+
+// 获取产品检测 pipeline 队列深度信息快照
+#define MEM_LOG(tag, fmt, ...) \
+    do { \
+        long _rss = getRSS_KB(); \
+        printf("[MEM] %-30s | VmRSS: %6ld MB | " fmt "\n", tag, _rss / 1024, ##__VA_ARGS__); \
+    } while(0)
+
+// 获取 GPU 显存使用情况（通过 nvidia-smi），返回所有 GPU 累加的已用/总量，单位 MB
+// 成功返回 0，失败返回 -1（此时 usedMB/totalMB 置为 -1）
+static inline int getGPUMemInfo(long &usedMB, long &totalMB)
+{
+    usedMB = -1;
+    totalMB = -1;
+    FILE *fp = popen("nvidia-smi --query-gpu=memory.used,memory.total --format=csv,noheader,nounits 2>/dev/null", "r");
+    if (!fp)
+    {
+        return -1;
+    }
+    char line[256];
+    long used = 0;
+    long total = 0;
+    int n = 0;
+    while (fgets(line, sizeof(line), fp))
+    {
+        long u = 0;
+        long t = 0;
+        if (sscanf(line, "%ld, %ld", &u, &t) == 2)
+        {
+            used += u;
+            total += t;
+            n++;
+        }
+    }
+    pclose(fp);
+    if (n == 0)
+    {
+        return -1;
+    }
+    usedMB = used;
+    totalMB = total;
+    return 0;
+}
+
+// 获取系统内存信息（/proc/meminfo），单位 KB
+// 成功返回 0，失败返回 -1（此时各值置为 -1）
+static inline int getSystemMemInfo(long &totalKB, long &freeKB, long &availableKB)
+{
+    totalKB = -1;
+    freeKB = -1;
+    availableKB = -1;
+    FILE *fp = fopen("/proc/meminfo", "r");
+    if (!fp)
+    {
+        return -1;
+    }
+    char line[256];
+    while (fgets(line, sizeof(line), fp))
+    {
+        long val = -1;
+        if (sscanf(line, "MemTotal: %ld kB", &val) == 1)
+        {
+            totalKB = val;
+        }
+        else if (sscanf(line, "MemFree: %ld kB", &val) == 1)
+        {
+            freeKB = val;
+        }
+        else if (sscanf(line, "MemAvailable: %ld kB", &val) == 1)
+        {
+            availableKB = val;
+        }
+    }
+    fclose(fp);
+    return 0;
+}
+
+// 系统 CPU 时间快照（/proc/stat 第一行 cpu 累计值），用于两次采样计算 CPU 使用率
+struct CPUStat
+{
+    unsigned long long user;
+    unsigned long long nice;
+    unsigned long long system;
+    unsigned long long idle;
+    unsigned long long iowait;
+    unsigned long long irq;
+    unsigned long long softirq;
+    unsigned long long steal;
+
+    CPUStat()
+    {
+        user = nice = system = idle = iowait = irq = softirq = steal = 0;
+    }
+    unsigned long long total() const
+    {
+        return user + nice + system + idle + iowait + irq + softirq + steal;
+    }
+    unsigned long long idleTotal() const
+    {
+        return idle + iowait;
+    }
+};
+
+// 读取系统 CPU 时间快照，成功返回 0，失败返回 -1
+static inline int getSystemCPUStat(CPUStat &stat)
+{
+    FILE *fp = fopen("/proc/stat", "r");
+    if (!fp)
+    {
+        return -1;
+    }
+    char line[256];
+    int ret = -1;
+    if (fgets(line, sizeof(line), fp))
+    {
+        // 格式: cpu  user nice system idle iowait irq softirq steal guest guest_nice
+        if (sscanf(line, "cpu %llu %llu %llu %llu %llu %llu %llu %llu",
+                   &stat.user, &stat.nice, &stat.system, &stat.idle,
+                   &stat.iowait, &stat.irq, &stat.softirq, &stat.steal) >= 4)
+        {
+            ret = 0;
+        }
+    }
+    fclose(fp);
+    return ret;
+}
+
+// 进程累计消耗的 CPU 时间（用户态 + 内核态），单位秒
+static inline double cpuTimeSec(const struct rusage &usage)
+{
+    return (double)usage.ru_utime.tv_sec + (double)usage.ru_utime.tv_usec / 1e6 +
+           (double)usage.ru_stime.tv_sec + (double)usage.ru_stime.tv_usec / 1e6;
+}
+
+#endif // _MEM_MONITOR_H_

example/SystemCommonDefine.h

@@ -46,6 +46,7 @@ struct SystemConfigParam
 
 	
 	std::string config_Root_Path;
+	std::string save_Img_Root_Path; // 检测结果图保存根目录
 
 	int Use_CPU_StartIdx;		  // 使用 CPU 的开始核
     int nWorkIdx;				  // 工位号
@@ -54,6 +55,7 @@ struct SystemConfigParam
 	{
 		Use_CPU_StartIdx = 0;
 		config_Root_Path = "";
+		save_Img_Root_Path = "/home/aidlux/BOE/FOG/Detsave/";
 		nWorkIdx = 0;
         for(int i=0; i<4; i++) GPU_Device_ID_List[i] = -1;
 	}

example/deal.cpp

@@ -3,10 +3,15 @@
 #include <sys/time.h>
 #include <sys/types.h>
 #include <sys/stat.h>
+#include <ctime>
+#include "MemMonitor.h"
 #include <unistd.h>
 #include <string>
+#include <algorithm>
 #include "CheckUtil.hpp"
 #include "LoadImage.hpp" // #include "LoadImage.h"
+#include "JsonConfig.h"
+#include "AI_Factory.h"
 
 bool check_strChannel(string strChannel){
 	if(strChannel == "XD" || strChannel == "XON") return false;
@@ -260,6 +265,9 @@ int deal::Init()
     // 初始化参数模块
     InitConfig();
 
+    // 设置标签检测模式标志，供 ChannelFuntonConfigJson 在解析 JSON 时使用
+    ChannelFuntonConfigJson::s_bDetTag = runConfig.det_Tag;
+
     printf("\n\n\n\n\n\n\n\n\n\n\n**********************************************************LoadCheckImgConfig\n");
     // 载入参数
     re = LoadCheckImgConfig();
@@ -833,7 +841,7 @@ void deal::ResultThread(int id)
             if (true)
             {
                 // 存图
-                std::string savepath = "/home/aidlux/BOE/FOG/Detsave/";
+                std::string savepath = m_system_param.save_Img_Root_Path;
                 saveImg(savepath, dealResult);
                 // saveImg("/home/aidlux/BOE/ResultImg/", dealResult);
             }
@@ -842,7 +850,7 @@ void deal::ResultThread(int id)
                 // 存 检测统计 日志
                 mutex_DetResult_.lock();
                 m_DetResult.print(false);
-                std::string strlog = "/home/aidlux/BOE/FOG/Detsave/";
+                std::string strlog = m_system_param.save_Img_Root_Path;
                 if (m_strCurDate != "")
                 {
                     strlog += m_strCurDate;
@@ -856,6 +864,10 @@ void deal::ResultThread(int id)
                 writeLog(strlog, m_DetResult.strlist);
                 mutex_DetResult_.unlock();
             }
+
+            // ✅ 递增已保存结果计数，并通知等待线程，立刻唤醒检查
+            m_nProductSavedResultCount.fetch_add(1);
+            m_productResultCV.notify_all();
         }
 
         usleep(5 * 1000);
@@ -930,6 +942,10 @@ bool deal::ReadSystemConfig(const std::string &strPath)
     m_system_param.Use_CPU_StartIdx = root["Use_CPU_StartIdx"].asInt();
     m_system_param.nWorkIdx = root["Detect_Work_Idx"].asInt();
     m_system_param.config_Root_Path = root["Config_Root_Path"].asString();
+    if (root.isMember("Save_Img_Root_Path"))
+    {
+        m_system_param.save_Img_Root_Path = root["Save_Img_Root_Path"].asString();
+    }
 
     for (int i = 0; i < 4; i++)
         m_system_param.GPU_Device_ID_List[i] = -1;
@@ -1289,6 +1305,15 @@ int deal::DetImg()
         return 0;
     }
 
+    // AI 检测独立测试：读取 in 图 -> 模型推理 -> 保存 out 图（不做其他流程）
+    if (runConfig.det_Type == RUNTYPE_RUN_File_AI_DetTest)
+    {
+        m_nReadThread_type = READ_THREAD_TYPE_READIMG;
+        m_nSystemCheckType = CHECK_TYPE_OFFLING;
+        TestAIDet();
+        return 0;
+    }
+
     m_nReadThread_type = READ_THREAD_TYPE_READIMG;
     m_nSystemCheckType = CHECK_TYPE_OFFLING;
 
@@ -1306,6 +1331,657 @@ int deal::DetImg()
     return 0;
 }
 
+/// @brief 监控模式：4 线程并发循环执行模型推理（压 GPU），并把当前设备资源消耗
+///        （CPU 内存 / GPU 显存）写入 monitor.log；正常（out 全黑）不写图，检测到残点
+///        （out 图至少一个像素不为全黑）时才保存对应 *_out.png 并退出监控。
+/// @return 0 表示成功，非 0 表示初始化失败
+int deal::MonitorRun()
+{
+    // 输入目录：优先使用命令行指定路径，否则使用默认路径
+    std::string strSrcRoot = runConfig.filePath;
+    if (strSrcRoot.empty())
+    {
+        strSrcRoot = "/home/aidlux/BOE/FOG/AI_Detect/";
+    }
+    else if (strSrcRoot.back() != '/')
+    {
+        strSrcRoot += "/";
+    }
+
+    std::shared_ptr<AIFactory> pAIFactory = AIFactory::GetInstance();
+    std::shared_ptr<AIModel_Base> pAI_Model = pAIFactory->AI_defect_NF;
+    if (!pAI_Model || !pAI_Model->m_bInitSuccess)
+    {
+        printf("MonitorRun: AI model (AI_defect_NF) not ready!\n");
+        return 1;
+    }
+
+    const int nModelInCh = pAI_Model->input_0.channel;
+    const int nModelInW = pAI_Model->input_0.width;
+    const int nModelInH = pAI_Model->input_0.height;
+
+    printf("MonitorRun: model input [c=%d w=%d h=%d] output [c=%d w=%d h=%d]\n",
+           nModelInCh, nModelInW, nModelInH,
+           pAI_Model->output_0.channel, pAI_Model->output_0.width, pAI_Model->output_0.height);
+
+    // 收集所有 *_in.png 文件
+    std::vector<std::string> inList;
+    try
+    {
+        if (!fs::exists(strSrcRoot))
+        {
+            printf("MonitorRun: src dir not exist: %s\n", strSrcRoot.c_str());
+            return 1;
+        }
+        for (const auto &entry : fs::recursive_directory_iterator(strSrcRoot))
+        {
+            if (!entry.is_regular_file())
+            {
+                continue;
+            }
+            std::string name = entry.path().filename().string();
+            const std::string suffix = "_in.png";
+            if (name.size() >= suffix.size() &&
+                name.compare(name.size() - suffix.size(), suffix.size(), suffix) == 0)
+            {
+                inList.push_back(entry.path().string());
+            }
+        }
+    }
+    catch (const std::exception &e)
+    {
+        printf("MonitorRun: traverse error %s\n", e.what());
+        return 1;
+    }
+
+    std::sort(inList.begin(), inList.end());
+    printf("MonitorRun: total %zu in images\n", inList.size());
+    if (inList.empty())
+    {
+        printf("MonitorRun: no *_in.png found\n");
+        return 1;
+    }
+
+    // 输出目录：输入目录去掉末尾 '/' 后追加 "_TEST/"（与 TestAIDet 保持一致）
+    std::string strDstRoot = strSrcRoot;
+    while (strDstRoot.size() > 1 && strDstRoot.back() == '/')
+    {
+        strDstRoot.pop_back();
+    }
+    strDstRoot += "_TEST/";
+    {
+        std::error_code ec;
+        fs::create_directories(strDstRoot, ec);
+    }
+
+    // 预处理成模型输入（只读一次，每轮复用），并计算对应输出路径（*_in.png -> *_out.png）
+    std::vector<std::string> okInList;
+    std::vector<cv::Mat> inImgs;
+    std::vector<std::string> outList;
+    okInList.reserve(inList.size());
+    inImgs.reserve(inList.size());
+    outList.reserve(inList.size());
+    for (size_t i = 0; i < inList.size(); i++)
+    {
+        cv::Mat inImg = cv::imread(inList[i], cv::IMREAD_UNCHANGED);
+        if (inImg.empty())
+        {
+            printf("MonitorRun: read fail: %s\n", inList[i].c_str());
+            continue;
+        }
+        if (inImg.channels() != nModelInCh)
+        {
+            cv::Mat tmp;
+            if (inImg.channels() >= 3)
+            {
+                cv::cvtColor(inImg, tmp, cv::COLOR_BGR2GRAY);
+            }
+            else
+            {
+                tmp = inImg.clone();
+            }
+            inImg = tmp;
+        }
+        if (inImg.type() != CV_8U)
+        {
+            inImg.convertTo(inImg, CV_8U);
+        }
+        if (inImg.cols != nModelInW || inImg.rows != nModelInH)
+        {
+            cv::resize(inImg, inImg, cv::Size(nModelInW, nModelInH));
+        }
+        okInList.push_back(inList[i]);
+        inImgs.push_back(inImg);
+
+        std::string rel = inList[i].substr(strSrcRoot.size());
+        const std::string inSuffix = "_in.png";
+        const std::string outSuffix = "_out.png";
+        size_t pos = rel.rfind(inSuffix);
+        if (pos != std::string::npos)
+        {
+            rel.replace(pos, inSuffix.size(), outSuffix);
+        }
+        outList.push_back(strDstRoot + rel);
+    }
+    if (inImgs.empty())
+    {
+        printf("MonitorRun: no valid in image\n");
+        return 1;
+    }
+
+    // 打开监控日志（追加写）
+    std::string strLogPath = strDstRoot + "monitor.log";
+    std::ofstream logFile(strLogPath, std::ios::app);
+    if (!logFile)
+    {
+        printf("MonitorRun: open log fail: %s\n", strLogPath.c_str());
+        return 1;
+    }
+
+    // 日志互斥锁（worker 线程与主线程都会写日志）
+    std::mutex logMutex;
+    auto logLine = [&logFile, &logMutex](const std::string &line)
+    {
+        std::lock_guard<std::mutex> lock(logMutex);
+        printf("%s\n", line.c_str());
+        fflush(stdout); // 强制刷新控制台，避免 stdout 块缓冲导致显示延迟
+        logFile << line << std::endl;
+    };
+
+    const int nThreadNum = 4; // 并发线程数（与模型 Stream_num 一致时并行度最高）
+    long nImageNum = (long)inImgs.size();
+
+    // 共享状态
+    std::atomic<long> nNextIdx{0};     // 原子分配任务索引，轮询所有图
+    std::atomic<long> nInferCount{0};  // 成功推理次数
+    std::atomic<long> nFailCount{0};   // 推理失败次数
+    std::atomic<bool> bHit{false};     // 是否命中非全黑 out
+    std::atomic<long> nHitIdx{-1};     // 命中的图索引
+    std::atomic<int>  nHitNonZero{0};  // 命中的非零像素数
+
+    std::mutex saveMutex; // 保护 out 文件写入
+
+    logLine("===== monitor start (4 threads loop, exit on first non-black out) =====");
+
+    // 启动 worker 线程：循环推理压 GPU
+    std::vector<std::thread> workers;
+    workers.reserve(nThreadNum);
+    for (int t = 0; t < nThreadNum; t++)
+    {
+        workers.emplace_back([&, t]()
+        {
+            while (!bHit.load(std::memory_order_relaxed))
+            {
+                long idx = nNextIdx.fetch_add(1, std::memory_order_relaxed) % nImageNum;
+
+                cv::Mat outImg;
+                int re = pAI_Model->AIDet(inImgs[(size_t)idx], outImg);
+                if (re != 0 || outImg.empty())
+                {
+                    nFailCount.fetch_add(1, std::memory_order_relaxed);
+                    continue;
+                }
+
+                nInferCount.fetch_add(1, std::memory_order_relaxed);
+
+                int nNonZero = cv::countNonZero(outImg);
+                if (nNonZero > 0)
+                {
+                    // 有残点才写图，然后退出
+                    bool expected = false;
+                    if (bHit.compare_exchange_strong(expected, true))
+                    {
+                        {
+                            std::lock_guard<std::mutex> lock(saveMutex);
+                            fs::path outParent = fs::path(outList[(size_t)idx]).parent_path();
+                            if (!outParent.empty())
+                            {
+                                std::error_code ec;
+                                fs::create_directories(outParent, ec);
+                            }
+                            cv::imwrite(outList[(size_t)idx], outImg);
+                        }
+                        nHitIdx.store(idx, std::memory_order_relaxed);
+                        nHitNonZero.store(nNonZero, std::memory_order_relaxed);
+                        logLine("[hit] thread " + std::to_string(t) + " out image " +
+                                okInList[(size_t)idx] + " non-zero pixels: " +
+                                std::to_string(nNonZero) + " -> " + outList[(size_t)idx]);
+                    }
+                    break;
+                }
+            }
+        });
+    }
+
+    // 主线程：资源监控（持续循环采样），直到命中退出
+    struct rusage usagePrev;
+    getrusage(RUSAGE_SELF, &usagePrev);
+    CPUStat cpuStatPrev;
+    getSystemCPUStat(cpuStatPrev);
+    auto wallPrev = std::chrono::steady_clock::now();
+
+    long nRound = 0;
+
+    // GPU 显存采样很慢（内部执行 nvidia-smi），降低采样频率避免阻塞主循环
+    long gpuUsedMB = -1;
+    long gpuTotalMB = -1;
+    const int nGPUSampleInterval = 10; // 每 10 轮（约 10 秒）才调用一次 nvidia-smi
+
+    while (!bHit.load(std::memory_order_relaxed))
+    {
+        nRound++;
+
+        // 1. 采集设备资源消耗
+        long rssKB = getRSS_KB();
+        struct rusage usage;
+        getrusage(RUSAGE_SELF, &usage);
+
+        long memTotalKB = -1;
+        long memFreeKB = -1;
+        long memAvailKB = -1;
+        getSystemMemInfo(memTotalKB, memFreeKB, memAvailKB);
+
+        CPUStat cpuStat;
+        getSystemCPUStat(cpuStat);
+
+        if ((nRound % nGPUSampleInterval) == 1)
+        {
+            getGPUMemInfo(gpuUsedMB, gpuTotalMB);
+        }
+
+        auto wallNow = std::chrono::steady_clock::now();
+
+        // 系统 CPU 使用率（两次采样之间）
+        double sysCpuUsage = -1.0;
+        unsigned long long totalDelta = cpuStat.total() - cpuStatPrev.total();
+        unsigned long long idleDelta = cpuStat.idleTotal() - cpuStatPrev.idleTotal();
+        if (totalDelta > 0)
+        {
+            sysCpuUsage = 100.0 * (1.0 - (double)idleDelta / (double)totalDelta);
+        }
+
+        // 进程 CPU 使用率（getrusage 的 CPU 时间 / 墙钟时间，多核下可能超过 100%）
+        double procCpuUsage = -1.0;
+        double wallSec = std::chrono::duration<double>(wallNow - wallPrev).count();
+        if (wallSec > 0)
+        {
+            procCpuUsage = 100.0 * (cpuTimeSec(usage) - cpuTimeSec(usagePrev)) / wallSec;
+        }
+
+        // 更新采样基准
+        usagePrev = usage;
+        cpuStatPrev = cpuStat;
+        wallPrev = wallNow;
+
+        time_t now = time(NULL);
+        struct tm tmv;
+        localtime_r(&now, &tmv);
+        char ts[64] = {0};
+        strftime(ts, sizeof(ts), "%Y-%m-%d %H:%M:%S", &tmv);
+
+        char buf[512];
+        snprintf(buf, sizeof(buf),
+                 "[round %ld] %s | infer=%ld fail=%ld | VmRSS: %ld MB | maxRSS: %ld MB | ProcCPU: %.1f%% | SysCPU: %.1f%% | Mem total %ld / free %ld / avail %ld MB | GPU: %ld/%ld MB",
+                 nRound, ts, nInferCount.load(), nFailCount.load(),
+                 rssKB / 1024, usage.ru_maxrss / 1024,
+                 procCpuUsage, sysCpuUsage,
+                 memTotalKB / 1024, memFreeKB / 1024, memAvailKB / 1024,
+                 gpuUsedMB, gpuTotalMB);
+        // 打印降频：每 10 轮打印一次，减少日志 IO/系统调用对 CPU 的争抢
+        if ((nRound % 10) == 1)
+        {
+            logLine(buf);
+        }
+    }
+
+    // 等待所有 worker 线程退出
+    for (auto &th : workers)
+    {
+        if (th.joinable())
+        {
+            th.join();
+        }
+    }
+
+    long hitIdx = nHitIdx.load();
+    logLine("===== monitor exit (hit image " +
+            (hitIdx >= 0 ? okInList[(size_t)hitIdx] : std::string("?")) +
+            ", non-zero pixels " + std::to_string(nHitNonZero.load()) + ") =====");
+
+    logFile.close();
+    return 0;
+}
+
+/// @brief AI 检测独立测试：遍历 /home/aidlux/BOE/FOG/AI_Detect/ 下所有 *_in.png，
+///        送入 AI_defect_NF 模型推理，将输出 *_out.png 保存到 /home/aidlux/BOE/FOG/AI_Detect_TEST/
+/// @return 0 表示成功
+/*int deal::TestAIDet()
+{
+    // 输入目录：优先使用命令行指定路径，否则使用默认路径
+    std::string strSrcRoot = runConfig.filePath;
+    if (strSrcRoot.empty())
+    {
+        strSrcRoot = "/home/aidlux/BOE/FOG/AI_Detect/";
+    }
+    else if (strSrcRoot.back() != '/')
+    {
+        strSrcRoot += "/";
+    }
+
+    // 输出目录：输入目录去掉末尾 '/' 后追加 "_TEST/"
+    // 默认输入 /home/aidlux/BOE/FOG/AI_Detect/ -> 输出 /home/aidlux/BOE/FOG/AI_Detect_TEST/
+    std::string strDstRoot = strSrcRoot;
+    while (strDstRoot.size() > 1 && strDstRoot.back() == '/')
+    {
+        strDstRoot.pop_back();
+    }
+    strDstRoot += "_TEST/";
+
+    if(runConfig.bmonitor)
+    {
+        // 监控模式：每 5 分钟执行一次模型推理，记录设备资源消耗到 log；
+        // 当某次推理输出 out 图至少一个像素不为全黑时退出监控。
+        return MonitorRun();
+    }
+
+    std::shared_ptr<AIFactory> pAIFactory = AIFactory::GetInstance();
+    std::shared_ptr<AIModel_Base> pAI_Model = pAIFactory->AI_defect_NF;
+    if (!pAI_Model || !pAI_Model->m_bInitSuccess)
+    {
+        printf("TestAIDet: AI model (AI_defect_NF) not ready!\n");
+        return 1;
+    }
+
+    printf("TestAIDet: model input [c=%d w=%d h=%d] output [c=%d w=%d h=%d]\n",
+           pAI_Model->input_0.channel, pAI_Model->input_0.width, pAI_Model->input_0.height,
+           pAI_Model->output_0.channel, pAI_Model->output_0.width, pAI_Model->output_0.height);
+
+    // 收集所有 *_in.png 文件
+    std::vector<std::string> inList;
+    try
+    {
+        if (!fs::exists(strSrcRoot))
+        {
+            printf("TestAIDet: src dir not exist: %s\n", strSrcRoot.c_str());
+            return 1;
+        }
+        for (const auto &entry : fs::recursive_directory_iterator(strSrcRoot))
+        {
+            if (!entry.is_regular_file())
+            {
+                continue;
+            }
+            std::string path = entry.path().string();
+            std::string name = entry.path().filename().string();
+            const std::string suffix = "_in.png";
+            if (name.size() >= suffix.size() &&
+                name.compare(name.size() - suffix.size(), suffix.size(), suffix) == 0)
+            {
+                inList.push_back(path);
+            }
+        }
+    }
+    catch (const std::exception &e)
+    {
+        printf("TestAIDet: traverse error %s\n", e.what());
+        return 1;
+    }
+
+    std::sort(inList.begin(), inList.end());
+    printf("TestAIDet: total %zu in images\n", inList.size());
+
+    const int nModelInCh = pAI_Model->input_0.channel;
+    const int nModelInW = pAI_Model->input_0.width;
+    const int nModelInH = pAI_Model->input_0.height;
+
+    int nOK = 0;
+    for (size_t i = 0; i < inList.size(); i++)
+    {
+        const std::string &inPath = inList[i];
+        cv::Mat inImg = cv::imread(inPath, cv::IMREAD_UNCHANGED);
+        if (inImg.empty())
+        {
+            printf("[%zu/%zu] read fail: %s\n", i + 1, inList.size(), inPath.c_str());
+            continue;
+        }
+
+        // 统一成模型要求的输入：单通道 8U、模型输入尺寸
+        if (inImg.channels() != nModelInCh)
+        {
+            cv::Mat tmp;
+            if (inImg.channels() >= 3)
+            {
+                cv::cvtColor(inImg, tmp, cv::COLOR_BGR2GRAY);
+            }
+            else
+            {
+                tmp = inImg.clone();
+            }
+            inImg = tmp;
+        }
+        if (inImg.type() != CV_8U)
+        {
+            inImg.convertTo(inImg, CV_8U);
+        }
+        if (inImg.cols != nModelInW || inImg.rows != nModelInH)
+        {
+            cv::resize(inImg, inImg, cv::Size(nModelInW, nModelInH));
+        }
+
+        cv::Mat outImg;
+        int re = pAI_Model->AIDet(inImg, outImg);
+        if (re != 0 || outImg.empty())
+        {
+            printf("[%zu/%zu] infer fail(%d): %s\n", i + 1, inList.size(), re, inPath.c_str());
+            continue;
+        }
+
+        // 输出 resize 回模型输入尺寸（与 AI_Detect_QX 保存的 out 图尺寸保持一致）
+        if (outImg.cols != nModelInW || outImg.rows != nModelInH)
+        {
+            cv::resize(outImg, outImg, cv::Size(nModelInW, nModelInH), 0, 0, 0);
+        }
+
+        // 相对路径，并保持子目录结构
+        std::string rel = inPath.substr(strSrcRoot.size());
+        std::string outRel = rel;
+        const std::string inSuffix = "_in.png";
+        const std::string outSuffix = "_out.png";
+        size_t pos = outRel.rfind(inSuffix);
+        if (pos != std::string::npos)
+        {
+            outRel.replace(pos, inSuffix.size(), outSuffix);
+        }
+        std::string outPath = strDstRoot + outRel;
+
+        // 创建输出父目录
+        fs::path outParent = fs::path(outPath).parent_path();
+        if (!outParent.empty())
+        {
+            std::error_code ec;
+            fs::create_directories(outParent, ec);
+        }
+
+        if (cv::imwrite(outPath, outImg))
+        {
+            nOK++;
+            printf("[%zu/%zu] ok: %s -> %s\n", i + 1, inList.size(), rel.c_str(), outRel.c_str());
+        }
+        else
+        {
+            printf("[%zu/%zu] save fail: %s\n", i + 1, inList.size(), outPath.c_str());
+        }
+    }
+
+    printf("TestAIDet: complete, success %d / %zu\n", nOK, inList.size());
+    return 0;
+}*/
+
+/// @brief AI 检测独立测试：遍历 /home/aidlux/BOE/FOG/AI_Detect/ 下所有 *_in.png，
+///        送入 AI_defect_NF 模型推理，将输出 *_out.png 保存到 /home/aidlux/BOE/FOG/AI_Detect_TEST/
+/// @return 0 表示成功
+int deal::TestAIDet()
+{
+    // 输入目录：优先使用命令行指定路径，否则使用默认路径
+    std::string strSrcRoot = runConfig.filePath;
+    if (strSrcRoot.empty())
+    {
+        strSrcRoot = "/home/aidlux/BOE/FOG/AI_Detect/";
+    }
+    else if (strSrcRoot.back() != '/')
+    {
+        strSrcRoot += "/";
+    }
+
+    // 输出目录：输入目录去掉末尾 '/' 后追加 "_TEST/"
+    // 默认输入 /home/aidlux/BOE/FOG/AI_Detect/ -> 输出 /home/aidlux/BOE/FOG/AI_Detect_TEST/
+    std::string strDstRoot = strSrcRoot;
+    while (strDstRoot.size() > 1 && strDstRoot.back() == '/')
+    {
+        strDstRoot.pop_back();
+    }
+    strDstRoot += "_TEST/";
+
+    if(runConfig.bmonitor)
+    {
+        // 监控模式：每 5 分钟执行一次模型推理，记录设备资源消耗到 log；
+        // 当某次推理输出 out 图至少一个像素不为全黑时退出监控。
+        return MonitorRun();
+    }
+
+    std::shared_ptr<AIFactory> pAIFactory = AIFactory::GetInstance();
+    std::shared_ptr<AIModel_Base> pAI_Model = pAIFactory->AI_defect_NF;
+    if (!pAI_Model || !pAI_Model->m_bInitSuccess)
+    {
+        printf("TestAIDet: AI model (AI_defect_NF) not ready!\n");
+        return 1;
+    }
+
+    printf("TestAIDet: model input [c=%d w=%d h=%d] output [c=%d w=%d h=%d]\n",
+           pAI_Model->input_0.channel, pAI_Model->input_0.width, pAI_Model->input_0.height,
+           pAI_Model->output_0.channel, pAI_Model->output_0.width, pAI_Model->output_0.height);
+
+    // 收集所有 *_in.png 文件
+    std::vector<std::string> inList;
+    try
+    {
+        if (!fs::exists(strSrcRoot))
+        {
+            printf("TestAIDet: src dir not exist: %s\n", strSrcRoot.c_str());
+            return 1;
+        }
+        for (const auto &entry : fs::recursive_directory_iterator(strSrcRoot))
+        {
+            if (!entry.is_regular_file())
+            {
+                continue;
+            }
+            std::string path = entry.path().string();
+            std::string name = entry.path().filename().string();
+            const std::string suffix = "_in.png";
+            if (name.size() >= suffix.size() &&
+                name.compare(name.size() - suffix.size(), suffix.size(), suffix) == 0)
+            {
+                inList.push_back(path);
+            }
+        }
+    }
+    catch (const std::exception &e)
+    {
+        printf("TestAIDet: traverse error %s\n", e.what());
+        return 1;
+    }
+
+    std::sort(inList.begin(), inList.end());
+    printf("TestAIDet: total %zu in images\n", inList.size());
+
+    const int nModelInCh = pAI_Model->input_0.channel;
+    const int nModelInW = pAI_Model->input_0.width;
+    const int nModelInH = pAI_Model->input_0.height;
+
+    int nOK = 0;
+    for (size_t i = 0; i < inList.size(); i++)
+    {
+        const std::string &inPath = inList[i];
+        cv::Mat inImg = cv::imread(inPath, cv::IMREAD_UNCHANGED);
+        if (inImg.empty())
+        {
+            printf("[%zu/%zu] read fail: %s\n", i + 1, inList.size(), inPath.c_str());
+            continue;
+        }
+
+        // 统一成模型要求的输入：单通道 8U、模型输入尺寸
+        if (inImg.channels() != nModelInCh)
+        {
+            cv::Mat tmp;
+            if (inImg.channels() >= 3)
+            {
+                cv::cvtColor(inImg, tmp, cv::COLOR_BGR2GRAY);
+            }
+            else
+            {
+                tmp = inImg.clone();
+            }
+            inImg = tmp;
+        }
+        if (inImg.type() != CV_8U)
+        {
+            inImg.convertTo(inImg, CV_8U);
+        }
+        if (inImg.cols != nModelInW || inImg.rows != nModelInH)
+        {
+            cv::resize(inImg, inImg, cv::Size(nModelInW, nModelInH));
+        }
+
+        cv::Mat outImg;
+        int re = pAI_Model->AIDet(inImg, outImg);
+        if (re != 0 || outImg.empty())
+        {
+            printf("[%zu/%zu] infer fail(%d): %s\n", i + 1, inList.size(), re, inPath.c_str());
+            continue;
+        }
+
+        // 输出 resize 回模型输入尺寸（与 AI_Detect_QX 保存的 out 图尺寸保持一致）
+        if (outImg.cols != nModelInW || outImg.rows != nModelInH)
+        {
+            cv::resize(outImg, outImg, cv::Size(nModelInW, nModelInH), 0, 0, 0);
+        }
+
+        // 相对路径，并保持子目录结构
+        std::string rel = inPath.substr(strSrcRoot.size());
+        std::string outRel = rel;
+        const std::string inSuffix = "_in.png";
+        const std::string outSuffix = "_out.png";
+        size_t pos = outRel.rfind(inSuffix);
+        if (pos != std::string::npos)
+        {
+            outRel.replace(pos, inSuffix.size(), outSuffix);
+        }
+        std::string outPath = strDstRoot + outRel;
+
+        // 创建输出父目录
+        fs::path outParent = fs::path(outPath).parent_path();
+        if (!outParent.empty())
+        {
+            std::error_code ec;
+            fs::create_directories(outParent, ec);
+        }
+
+        if (cv::imwrite(outPath, outImg))
+        {
+            nOK++;
+            printf("[%zu/%zu] ok: %s -> %s\n", i + 1, inList.size(), rel.c_str(), outRel.c_str());
+        }
+        else
+        {
+            printf("[%zu/%zu] save fail: %s\n", i + 1, inList.size(), outPath.c_str());
+        }
+    }
+
+    printf("TestAIDet: complete, success %d / %zu\n", nOK, inList.size());
+    return 0;
+}
+
 /// @brief 检查文件图片
 /// @detail 后续实现具体功能
 /// @return 0 表示成功
@@ -1341,11 +2017,105 @@ int deal::CheckFileImg()
         products[i].print("产品" + std::to_string(i + 1));
     }
     // getchar();
+    printMemUsage("CheckFileImg-开始处理", "");
     for (size_t i = 0; i < products.size(); i++)
     {
+        char buf[128];
+        snprintf(buf, sizeof(buf), "产品[%zu/%zu] %s 处理前", i+1, products.size(), products[i].product_id.c_str());
+        printMemUsage(buf, "");
         
         CheckProduct(products[i]); // 处理单个产品
+        
+        // ✅ 等待当前产品的所有检测结果保存完成，再处理下一个产品
+        //     使用条件变量：ResultThread 每保存完一个结果就 notify_all，立刻唤醒检查
+        printf("\n[同步] 等待产品 %s 检测结果保存完成...\n", products[i].product_id.c_str());
+        {
+            std::unique_lock<std::mutex> lock(mutex_Result_list);
+            int waitCount = 0;
+            while (true)
+            {
+                bool done = m_productResultCV.wait_for(lock, std::chrono::milliseconds(5000), [this]() {
+                    return m_nProductSavedResultCount.load() >= m_nProductSuccessCount.load() 
+                           && m_Result_list.empty();
+                });
+                if (done) break;
+                waitCount++;
+                printf("[同步] 等待中... 已保存: %d/%d (成功:%d 失败:%d), 结果队列: %s\n", 
+                       m_nProductSavedResultCount.load(), 
+                       m_nProductSuccessCount.load() + m_nProductFailCount.load(),
+                       m_nProductSuccessCount.load(), m_nProductFailCount.load(),
+                       m_Result_list.empty() ? "空" : "非空");
+            }
+        }
+        printf("[同步] 产品 %s 所有结果已保存完成 (saved=%d, success=%d)\n", 
+               products[i].product_id.c_str(), 
+               m_nProductSavedResultCount.load(), 
+               m_nProductSuccessCount.load());
+        
+        // ✅ 释放资源：清理图像数据，释放内存
+        {
+            // 清理图像信息列表（已在 CheckProduct 开头清理过，这里做双重保障）
+            for (auto &info : m_ImageInfoList)
+            {
+                if (!info.img.empty())
+                {
+                    info.img.release();
+                }
+            }
+            m_ImageInfoList.clear();
+            std::vector<JC_IMAGE_INFO_>().swap(m_ImageInfoList); // 强制释放 vector 内存
+        }
+        
+        snprintf(buf, sizeof(buf), "产品[%zu/%zu] %s 资源释放后", i+1, products.size(), products[i].product_id.c_str());
+        printMemUsage(buf, "");
+    }
+    // ✅ 修复：所有产品处理完成后，通知后台线程退出
+    printf("\n>>> 所有产品处理完成，正在停止后台线程...\n");
+    
+    // 1. 通知产品读图线程退出
+    m_bProductReadExit = true;
+    m_productTaskCV.notify_all();
+    m_checkNotifyCV.notify_all();
+    
+    // 2. 通知结果处理线程退出
+    m_bExit = true;
+    cv.notify_all();
+    
+    // 3. 等待产品读图线程结束
+    for (auto &thread : m_productReadThreads)
+    {
+        if (thread && thread->joinable())
+        {
+            thread->join();
+        }
     }
+    m_productReadThreads.clear();
+    
+    // 4. 等待结果处理线程结束
+    for (auto &thread : ptr_ResultthreadList)
+    {
+        if (thread && thread->joinable())
+        {
+            thread->join();
+        }
+    }
+    ptr_ResultthreadList.clear();
+    
+    // 5. 等待 GetResult 线程结束
+    if (ptr_GetResultthread && ptr_GetResultthread->joinable())
+    {
+        ptr_GetResultthread->join();
+    }
+    ptr_GetResultthread.reset();
+    
+    // 6. 等待 DealImg 线程结束（如果存在）
+    if (ptr_DealImgthread && ptr_DealImgthread->joinable())
+    {
+        ptr_DealImgthread->join();
+    }
+    ptr_DealImgthread.reset();
+    
+    printf(">>> 所有后台线程已停止，程序退出.\n");
     printf("\n");
 
     // TODO: 后续实现具体功能
@@ -1383,9 +2153,15 @@ int deal::CheckProduct(const LoadProductImages &product)
     m_nProductPendingTasks = 0;
     m_nProductSuccessCount = 0;
     m_nProductFailCount = 0;
+    m_nProductSavedResultCount = 0;  // 重置已保存结果计数
     m_ProductFailedList.clear();
     m_nCheckNotifiedCount = 0;
 
+    // ✅ 修复内存泄漏：清理上一个产品累积的原始图像数据（cv::Mat）
+    // 注意：m_DetResult 是统计结果（轻量字符串/数字），不清除，跨产品累积
+    m_ImageInfoList.clear();
+    m_productIdList.clear();
+
     // 3. 打印产品信息
     printf("\n");
     printf("╔══════════════════════════════════════════════════════════\n");

example/deal.h

@@ -138,6 +138,7 @@ enum RUNTYPE_
 	RUNTYPE_RUN_File_AI_TEST,
 	RUNTYPE_RUN_File_Edge_Test_Save_Qx,
 	RUNTYPE_RUN_File_Edge_Test_Save_ALL,
+	RUNTYPE_RUN_File_AI_DetTest, // AI 检测独立测试：读 in 图 -> 推理 -> 存 out 图
 };
 // 程序运行模式
 enum Process_Run_Type
@@ -545,7 +546,9 @@ struct DealRunConfig
 	Process_Run_Type run_Type = Process_Run_Detect; // 运行模式
 	RUNTYPE_ det_Type = RUNTYPE_RUN_Pre;			// 处理
 	std::string filePath = "";						// 处理文件路径 如果为空 处理当前的单张图片
+	bool det_Tag = false;							// 是否检测标签
 	bool bDebugSaveImg = false;						// 是否中间图片图片
+	bool bmonitor = false;							// 是否监控
 	int m_nTestNum = 999999;
 	void print(std::string str)
 	{
@@ -632,6 +635,8 @@ private:
 	int ReJson();
 	int ReJson_ALL();
 	int DetImg();																					 // 套图复测
+	int TestAIDet(); // AI 检测独立测试：读取 in 图送入模型，保存 out 图
+	int MonitorRun();																 // 监控模式：每 5 分钟推理一次，记录资源消耗，遇非全黑 out 退出
 	int ReJson_product(std::vector<Json_Det_Path> &jsonList, std::string strProductName);			 // 套图复测
 	int CheckFileImg();																				 ///< 检查文件图片（新增）
 	int CheckProduct(const LoadProductImages &product);												 ///< 检查单个产品（新增）
@@ -799,6 +804,8 @@ public:
 	std::mutex m_checkNotifyMutex;					   // 通知队列锁
 	std::condition_variable m_checkNotifyCV;		   // 通知条件变量
 	std::atomic<int> m_nCheckNotifiedCount{0};		   // 已通知数量
+	std::atomic<int> m_nProductSavedResultCount{0};    // 已保存结果数量（用于产品间同步等待）
+	std::condition_variable m_productResultCV;          // 产品结果保存通知CV（每次save后notify）
 
 	std::set<std::string> m_productIdList; // 产品ID列表
 };

example/test_example.cpp

@@ -188,6 +188,14 @@ int main(int argc, char *argv[])
 					return 0;
 				}
 			}
+			else if (string(argv[i]) == "-testAIDet")
+			{
+				test.runConfig.det_Type = RUNTYPE_RUN_File_AI_DetTest;
+				if (argc == 3)
+				{
+					test.runConfig.filePath = string(argv[i + 1]);
+				}
+			}
 			else if (string(argv[i]) == "-feqx" && i + 1 < argc)
 			{
 				test.runConfig.det_Type = RUNTYPE_RUN_File_Edge_Test_Save_Qx;
@@ -228,11 +236,20 @@ int main(int argc, char *argv[])
 					return 0;
 				}
 			}
-
+			else if (string(argv[i]) == "-tag" && i + 1 < argc)
+			{
+				test.runConfig.det_Type = RUNTYPE_RUN_File;
+				test.runConfig.filePath = string(argv[i + 1]);
+				test.runConfig.det_Tag = true;
+			}
 			if (string(argv[i]) == "-s")
 			{
 				test.runConfig.bDebugSaveImg = true;
 			}
+			if (string(argv[i]) == "-m")
+			{
+				test.runConfig.bmonitor = true;
+			}
 			if (string(argv[i]) == "-n" && i + 1 < argc)
 			{
 				std::string strnum = string(argv[i + 1]);
@@ -245,10 +262,10 @@ int main(int argc, char *argv[])
 	signal(SIGINT, handler);
 	test.start();
 
-	while (true)
-	{
-		usleep(10 * 1000);
-	}
+	// while (true)
+	// {
+	// 	usleep(10 * 1000);
+	// }
 
 	return 0;
 }