修正像素切分中的bug

ai_matrix/Config/Config.cpp

@@ -70,6 +70,8 @@ namespace ai_matrix
             this->identifyConfig_.iTargetMinWidth = config_["identify"]["target_min_width"].as<int>();
             this->identifyConfig_.iTargetMinY = config_["identify"]["target_min_y"].as<int>();
             this->identifyConfig_.iMaxIdentifyFrame = config_["identify"]["max_identify_frame"].as<int>();
+            this->identifyConfig_.iMaxContainerSpaceX = config_["identify"]["max_container_space_x"].as<int>();
+            this->identifyConfig_.iMaxContainerSpaceY = config_["identify"]["max_container_space_y"].as<int>();
 
             // websocket server 服务端参数
             this->wSocketConfig_.bIsUse = config_["wsocket_server"]["is_use"].as<bool>();

ai_matrix/Config/Config.h

@@ -88,6 +88,10 @@ namespace ai_matrix
         int iTargetMinY;
         // 单次识别最大帧数
         int iMaxIdentifyFrame;
+        // 两个箱子的箱角最大差值X
+        int iMaxContainerSpaceX;
+        // 两个箱子的箱角最大差值Y
+        int iMaxContainerSpaceY;
     };
 
     // websocket_server 的服务端参数

app/config/config.yaml

@@ -1,7 +1,7 @@
 # 基础控制参数
 base:
   # 股道名称
-  track_name: "1"
+  track_name: "1门"
   # 测试模式
   test_model: false
   # Api 监听端口
@@ -24,43 +24,63 @@ log:
 
 # 数据源参数
 data_source:
-  url: "./vedio/buertai2.mp4"
+  - # 顶部摄像头（必须顶部摄像头）
-  # 跳帧数
+    url: "./videos/buertai2.mp4"
-  skip_interval: 3
+    # 跳帧数
-  # 行驶方向  0-自动识别 1-向左 2-向右 （与“首位信息”成对存在，形成例如向左就编号在前，向右就属性在前的对应）
+    skip_interval: 3
-  direction: 0
+    # 识别区域
-  # 0-向左编号在前 1-向左属性在前 (向右行驶的情况：2-向右编号在前 3-向右属性在前)
+    identify_areas: [120, 0, 1800, 1080]
-  left_first: 0
+    # 切箱方式
-  # (向左行驶的情况：0-向左编号在前 1-向左属性在前) 2-向右编号在前 3-向右属性在前
+    divide_mode: "corner" #[corner, pixel]
-  right_first: 3
+    # 汽车行进方向
-  # 识别区域
+    run_direction: "down" #[up, down, left, right]
-  identify_areas: [120, 0, 1800, 1080]
+  - # 侧部摄像头
+    url: "./videos/buertai2.mp4"
+    # 跳帧数
+    skip_interval: 3
+    # 识别区域
+    identify_areas: [ 120, 0, 1800, 1080 ]
+    # 切箱方式
+    divide_mode: "pixel" #[corner, pixel]
+    # 汽车行进方向
+    run_direction: "right" #[up, down, left, right]
+  - # 侧边摄像头
+    url: "./videos/buertai2.mp4"
+    # 跳帧数
+    skip_interval: 3
+    # 识别区域
+    identify_areas: [ 120, 0, 1800, 1080 ]
+    # 切箱方式
+    divide_mode: "pixel" #[corner, pixel]
+    # 汽车行进方向
+    run_direction: "left" #[up, down, left, right]
 
 # 识别参数
 identify:
-  # 运行方式
+  # 切箱方式
-  run_mode: "always" #[always; command]
+  divide_mode: "corner" #[corner, pixel]
-  # 是否开启动态检测
-  need_move_detect_flag: true
-  # 识别方向  [LEFT,RIGHT,ALL]
-  identify_direction: "LEFT"
   # 大框帧跨度（比一个大框从出现到消失的跨度稍大一点, 跟跳帧有关系）
-  partition_frame_span: 20
+  partition_frame_span: 0
-  # 大框帧跨度的位置像素差异
+  # 顶部Y轴大框帧跨度的位置像素差异
-  split_frame_span_px: 200
+  top_y_split_span_px: 200
+  # 侧边X轴大框帧跨度的位置像素差异
+  side_x_split_span_px: 600
   # 每帧大框位置差异最小值 (持续小于此值，则可能停车)
   chkstop_px: 15
   # 持续X次续位置差异小于gc_chkstop_px，则判断为停车。
   chkstop_count: 10
   # 过滤最小大框高度（不需要的话就写个很小的值）
-  num_frame_height: 150
+  target_min_height: 95
-  pro_frame_height: 120
+  # 过滤最小大框宽度（不需要的话就写个很小的值）
-  # 过滤最大框宽度（不需要的话就写个很大的值）
+  target_min_width: 50
-  space_frame_width: 500
+  # 过滤低于指定Y轴的大框
-  # 是否识别车头
+  target_min_y: 550
-  train_heard_detect: false
+  # 单次识别最大帧数
-  # 是否识别集装箱
+  max_identify_frame: 600
-  container_detect: false
+  # 两个箱子的箱角最大差值X
+  max_container_space_x: 600
+  # 两个箱子的箱角最大差值Y
+  max_container_space_y: 500
 
 # websocket_server 的服务端参数
 wsocket_server:

app/config/matrix.yaml

@@ -1,44 +1,56 @@
-#use_deviceid:
-#  #engineid: deviceid
-#  0: 0
-
 #engine实例
 engines:
   ApiEngine: 0
+  DeleteExpiredFolderEngine: 0
+  WSServerEngine: 0
   VideoEngine: 0
+  VideoEngine: 1
+  VideoEngine: 2
   VideoDecodeEngine: 0
-  MoveEngine: 0
+  VideoDecodeEngine: 1
-  SaveMoveImageEngine: 0
+  VideoDecodeEngine: 2
+  ControlEngine: 0
+#  SaveMoveImageEngine: 0
   SaveMoveInfoEngine: 0
-  TrainStep1DataReadEngine: 0
+  ContainerStep1InferenceEngine: 0
-  TrainStep1InferenceEngine: 0
+  CornerInferenceEngine: 0
-  TrainStep1FilterEngine: 0
+  Step1MergeEngine: 0
-  TrainDivideEngine: 0
+  ContainerStep2InferenceEngine: 0
-  TrainStep2DataReadEngine: 0
+  ContainerCharacterConversionEngine: 0
-  TrainStep2InferenceEngine: 0
+  ContainerDivideEngine: 0
-  TrainCharacterConversionEngine: 0
+  ContainerDivideEngine: 1
+  ContainerDivideEngine: 2
   SelectBestEngine: 0
   SaveResultCSVEngine: 0
-  ToHttpSrvEngine: 0
   ToMinioSrvEngine: 0
   SaveDebugImageEngine: 0
 
 #engine连接
 connects:
+  WSServerEngine_0_0: "ControlEngine_0_0 1024"
   VideoEngine_0_0: "VideoDecodeEngine_0_0 1024"
-  VideoDecodeEngine_0_0: "MoveEngine_0_0 1024"
+  VideoEngine_1_0: "VideoDecodeEngine_1_0 1024"
-  MoveEngine_0_0: "SaveMoveImageEngine_0_0 1024"
+  VideoEngine_2_0: "VideoDecodeEngine_2_0 1024"
-  MoveEngine_0_1: "SaveMoveInfoEngine_0_0 1024"
+  VideoDecodeEngine_0_0: "ControlEngine_0_1 1024"
-  SaveMoveImageEngine_0_0: "TrainStep1DataReadEngine_0_0 1024"
+  VideoDecodeEngine_1_0: "ControlEngine_0_1 1024"
-  TrainStep1DataReadEngine_0_0: "TrainStep1InferenceEngine_0_0 1024"
+  VideoDecodeEngine_2_0: "ControlEngine_0_1 1024"
-  TrainStep1InferenceEngine_0_0: "TrainStep1FilterEngine_0_0 1024"
+  ControlEngine_0_0: "WSServerEngine_0_0 1024"
-  TrainStep1FilterEngine_0_0: "TrainDivideEngine_0_0 1024"
+  ControlEngine_0_1: "ContainerStep1InferenceEngine_0_0 1024"
-  TrainDivideEngine_0_0: "TrainStep2DataReadEngine_0_0 1024"
+  ControlEngine_0_2: "CornerInferenceEngine_0_0 1024"
-  TrainStep2DataReadEngine_0_0: "TrainStep2InferenceEngine_0_0 1024"
+  ControlEngine_0_3: "SaveMoveInfoEngine_0_0 1024"
-  TrainStep2InferenceEngine_0_0: "TrainCharacterConversionEngine_0_0 1024"
+ ControlEngine_0_4: "SaveMoveImageEngine_0_0 1024"
-  TrainCharacterConversionEngine_0_0: "SelectBestEngine_0_0 1024"
+  ContainerStep1InferenceEngine_0_0: "Step1MergeEngine_0_0 1024"
-  TrainCharacterConversionEngine_0_1: "SaveDebugImageEngine_0_0 1024"
+  CornerInferenceEngine_0_0: "Step1MergeEngine_0_1 1024"
-  SelectBestEngine_0_0: "SaveResultCSVEngine_0_0 1024"
+  Step1MergeEngine_0_0: "ContainerStep2InferenceEngine_0_0 1024"
-  SaveResultCSVEngine_0_0: "ToHttpSrvEngine_0_0 1024"
+  ContainerStep2InferenceEngine_0_0: "ContainerCharacterConversionEngine_0_0 1024"
-  SaveResultCSVEngine_0_1: "ToMinioSrvEngine_0_0 1024"
+  ContainerCharacterConversionEngine_0_0: "ContainerDivideEngine_0_0 1024"
+  ContainerCharacterConversionEngine_0_1: "ContainerDivideEngine_1_0 1024"
+  ContainerCharacterConversionEngine_0_2: "ContainerDivideEngine_2_0 1024"
+  ContainerCharacterConversionEngine_0_3: "SaveDebugImageEngine_0_0 1024"
+  ContainerDivideEngine_0_0: "SelectBestEngine_0_0 1024"
+  ContainerDivideEngine_1_0: "SelectBestEngine_0_0 1024"
+  ContainerDivideEngine_2_0: "SelectBestEngine_0_0 1024"
+  SelectBestEngine_0_0: "WSServerEngine_0_0 1024"
+  SelectBestEngine_0_1: "ToMinioSrvEngine_0_0 1024"
+  SelectBestEngine_0_2: "SaveResultCSVEngine_0_0 1024"
 

app/config/model_config.yaml

@@ -39,9 +39,17 @@ train_step2_model:
 container_step1_model:
   model_path: "./model/container_step1/con1.engine"
   score_threshold: 0.6
-  class: []
+  class: ["CONTAINERNUM","CONTAINERNUM_REVERSE"] # ,"SPACE_T","SPACENX_T","SPACEG_T","SPACEP_T","SPACEP_T"
 # 集装箱字符识别
 container_step2_model:
   model_path: "./model/container_step2/con2.engine"
   score_threshold: 0.7
-  class: []
+  class: ["0","1","2","3","4","5","6","7","8","9",
+          "A","B","C","D","E","F","G","H","I","J","K","L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V","W", "X", "Y", "Z",
+          "change", "load", "m", "self", "t", "volume", "meter",")", "(", "?", "-"
+  ]
+# 集装箱箱角识别
+corner_model:
+  model_path: "./model/corner/corner.engine"
+  score_threshold: 0.7
+  class: ["CNTR_CORNER","CNTR_CORNER_TOP"]

common/CommonStruct.h

@@ -380,4 +380,22 @@ typedef struct
 
 } DetectResultData;
 
+
+// 存图数据
+typedef struct
+{
+    // 数据来源标识
+    int iDataSource;
+    std::string strDetectDate;
+    std::string strDetectTime;
+    // 帧序号
+    int iFrameId;
+    // 箱号大框以及对应小框的集合
+    Step2ResultData step2ResultData;
+    // 箱角大框
+    std::vector<Step2ResultData> vecCornerResultData;
+    // 图片
+    cv::Mat cvImage;
+} SaveDebugImgData;
+
 #endif //TRAIN_COMMONSTRUCT_H

config/config.yaml

@@ -77,6 +77,10 @@ identify:
   target_min_y: 550
   # 单次识别最大帧数
   max_identify_frame: 600
+  # 两个箱子的箱角最大差值X
+  max_container_space_x: 600
+  # 两个箱子的箱角最大差值Y
+  max_container_space_y: 500
 
 # websocket_server 的服务端参数
 wsocket_server:

config/matrix.yaml

@@ -10,7 +10,7 @@ engines:
   VideoDecodeEngine: 1
   VideoDecodeEngine: 2
   ControlEngine: 0
-  SaveMoveImageEngine: 0
+#  SaveMoveImageEngine: 0
   SaveMoveInfoEngine: 0
   ContainerStep1InferenceEngine: 0
   CornerInferenceEngine: 0

engine/ControlEngine/ControlEngine.cpp

@@ -116,7 +116,7 @@ void ControlEngine::detectControl(std::shared_ptr<std::string> pWSServerOrder)
 
     switch (jvOrder["commandType"].asInt()) {
         case IDENTIFY_START:
-            if (g_identify_type == 1)
+            if (g_identify_type == IDENTIFY_START)
             {
                 std::string msg = "当前正在识别，无需重复发送识别信号";
                 LogWarn << msg;
@@ -245,12 +245,12 @@ APP_ERROR ControlEngine::Process()
             outputQueMap_[strPort3_]->push(std::static_pointer_cast<void>(pVDetectInfo), true);
 
             // 存图
-//            std::shared_ptr<SaveImgData> pSaveImgData = std::make_shared<SaveImgData>();
+            std::shared_ptr<SaveImgData> pSaveImgData = std::make_shared<SaveImgData>();
-//            pSaveImgData->strFilePath = strFilePath;
+            pSaveImgData->strFilePath = strFilePath;
-//            pSaveImgData->strFileName = std::to_string(this->mapDetectNO_[pProcessData->iDataSource]) + "_" + std::to_string(pProcessData->iDataSource) + ".jpg";
+            pSaveImgData->strFileName = std::to_string(this->mapDetectNO_[pProcessData->iDataSource]) + "_" + std::to_string(pProcessData->iDataSource) + ".jpg";
-//            pSaveImgData->cvImage = image;
+            pSaveImgData->cvImage = image;
-//            pSaveImgData->bIsEnd = pProcessData->bIsEnd;
+            pSaveImgData->bIsEnd = pProcessData->bIsEnd;
-//            outputQueMap_[strPort4_]->push(std::static_pointer_cast<void>(pSaveImgData), true);
+            outputQueMap_[strPort4_]->push(std::static_pointer_cast<void>(pSaveImgData), true);
 
             this->mapDetectNO_[pProcessData->iDataSource]++;
         }

engine/DataUploadEngine/ToMinioSrvEngine.cpp

@@ -166,7 +166,7 @@ APP_ERROR ToMinioSrvEngine::Process()
                            strMinIoPath,
                            strLocalPath))
             {
-                LogWarn << "数据上传失败！ -- " << this->baseConfig_.strDebugResultPath;
+                LogWarn << "数据上传失败！ -- " << strLocalPath;
             }
             else
             {

engine/MergeEngine/Step1MergeEngine.cpp

@@ -1 +1 @@
-#include "Step1MergeEngine.h"

using namespace ai_matrix;

namespace
{
    //按照x坐标排列
    bool CompareX(const SingleData &a, const SingleData &b)
    {
        return a.fLTX < b.fLTX;
    }
}

Step1MergeEngine::Step1MergeEngine() {}

Step1MergeEngine::~Step1MergeEngine() {}

APP_ERROR Step1MergeEngine::Init()
{
    strPort0_ = engineName_ + "_" + std::to_string(engineId_) + "_0";
    strPort1_ = engineName_ + "_" + std::to_string(engineId_) + "_1";

    this->baseConfig_ = Config::getins()->getBaseConfig();
    this->identifyConfig_ = Config::getins()->getIdentifyConfig();

    this->multiTypeQueue_ = new ai_matrix::MultiTypeQueue(2);
    LogInfo << "MergeEngine Init ok";
    return APP_ERR_OK;
}

APP_ERROR Step1MergeEngine::DeInit()
{
    LogInfo << "MergeEngine DeInit ok";
    return APP_ERR_OK;
}

APP_ERROR Step1MergeEngine::Process()
{
    int iRet = APP_ERR_OK;
    while (!isStop_)
    {
        std::shared_ptr<void> pVoidData0 = nullptr;
        inputQueMap_[strPort0_]->pop(pVoidData0);

        std::shared_ptr<void> pVoidData1 = nullptr;
        inputQueMap_[strPort1_]->pop(pVoidData1);

        if (nullptr == pVoidData0 && nullptr == pVoidData1)
        {
            usleep(1000); //1ms
            continue;
        }

        if (pVoidData0)
        {
            this->multiTypeQueue_->PushData(0, pVoidData0);
        }
        if (pVoidData1)
        {
            this->multiTypeQueue_->PushData(1, pVoidData1);
        }

        if (!this->multiTypeQueue_->PopAllData(pVoidData0, pVoidData1))
        {
            usleep(1000); //1ms
            continue;
        }

        std::shared_ptr<InferenceResultData> pInferenceResultData =
                std::static_pointer_cast<InferenceResultData>(pVoidData0);
        std::shared_ptr<InferenceResultData> pInferenceResultData_container =
                std::static_pointer_cast<InferenceResultData>(pVoidData1);

        pInferenceResultData->vecSingleData = pInferenceResultData_container->vecSingleData;

        std::sort(pInferenceResultData->vecSingleData.begin(),
                  pInferenceResultData->vecSingleData.end(),
                  CompareX);
        if (pInferenceResultData->singleData.fScore > 0.0f)
        {
            // 箱号
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " 数据源:" << pInferenceResultData->iDataSource
//                     << " --iClassId:" << pInferenceResultData->singleData.iClassId
//                     << " confidence=" << pInferenceResultData->singleData.fScore
//                     << " lx=" << pInferenceResultData->singleData.fLTX
//                     << " ly=" << pInferenceResultData->singleData.fLTY
//                     << " rx=" << pInferenceResultData->singleData.fRBX
//                     << " ry=" << pInferenceResultData->singleData.fRBY
//                     << " clear:" << pInferenceResultData->singleData.fClear;
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
        }

        // 箱角
        for (const auto & it_result : pInferenceResultData->vecSingleData)
        {
            LogDebug << " 帧:" << pInferenceResultData->iFrameId
                     << " 数据源:" << pInferenceResultData->iDataSource
                     << " --iClassId:" << it_result.iClassId
                     << " confidence=" << it_result.fScore
                     << " lx=" << it_result.fLTX
                     << " ly=" << it_result.fLTY
                     << " rx=" << it_result.fRBX
                     << " ry=" << it_result.fRBY;
        }

        outputQueMap_[strPort0_]->push(std::static_pointer_cast<void>(pInferenceResultData), true);
    }
    return APP_ERR_OK;
}
+#include "Step1MergeEngine.h"

using namespace ai_matrix;

namespace
{
    //按照x坐标排列
    bool CompareX(const SingleData &a, const SingleData &b)
    {
        return a.fLTX < b.fLTX;
    }
}

Step1MergeEngine::Step1MergeEngine() {}

Step1MergeEngine::~Step1MergeEngine() {}

APP_ERROR Step1MergeEngine::Init()
{
    strPort0_ = engineName_ + "_" + std::to_string(engineId_) + "_0";
    strPort1_ = engineName_ + "_" + std::to_string(engineId_) + "_1";

    this->baseConfig_ = Config::getins()->getBaseConfig();
    this->identifyConfig_ = Config::getins()->getIdentifyConfig();

    this->multiTypeQueue_ = new ai_matrix::MultiTypeQueue(2);
    LogInfo << "MergeEngine Init ok";
    return APP_ERR_OK;
}

APP_ERROR Step1MergeEngine::DeInit()
{
    LogInfo << "MergeEngine DeInit ok";
    return APP_ERR_OK;
}

APP_ERROR Step1MergeEngine::Process()
{
    int iRet = APP_ERR_OK;
    while (!isStop_)
    {
        std::shared_ptr<void> pVoidData0 = nullptr;
        inputQueMap_[strPort0_]->pop(pVoidData0);

        std::shared_ptr<void> pVoidData1 = nullptr;
        inputQueMap_[strPort1_]->pop(pVoidData1);

        if (nullptr == pVoidData0 && nullptr == pVoidData1)
        {
            usleep(1000); //1ms
            continue;
        }

        if (pVoidData0)
        {
            this->multiTypeQueue_->PushData(0, pVoidData0);
        }
        if (pVoidData1)
        {
            this->multiTypeQueue_->PushData(1, pVoidData1);
        }

        if (!this->multiTypeQueue_->PopAllData(pVoidData0, pVoidData1))
        {
            usleep(1000); //1ms
            continue;
        }

        std::shared_ptr<InferenceResultData> pInferenceResultData =
                std::static_pointer_cast<InferenceResultData>(pVoidData0);
        std::shared_ptr<InferenceResultData> pInferenceResultData_container =
                std::static_pointer_cast<InferenceResultData>(pVoidData1);

        pInferenceResultData->vecSingleData = pInferenceResultData_container->vecSingleData;

        std::sort(pInferenceResultData->vecSingleData.begin(),
                  pInferenceResultData->vecSingleData.end(),
                  CompareX);
        if (pInferenceResultData->singleData.fScore > 0.0f)
        {
            // 箱号
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " 数据源:" << pInferenceResultData->iDataSource
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " --iClassId:" << pInferenceResultData->singleData.iClassId
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " confidence=" << pInferenceResultData->singleData.fScore
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " lx=" << pInferenceResultData->singleData.fLTX
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " ly=" << pInferenceResultData->singleData.fLTY
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " rx=" << pInferenceResultData->singleData.fRBX
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " ry=" << pInferenceResultData->singleData.fRBY
//            LogDebug << " 帧:" << pInferenceResultData->iFrameId
//                     << " clear:" << pInferenceResultData->singleData.fClear;
        }

        // 箱角
        for (const auto & it_result : pInferenceResultData->vecSingleData)
        {
            LogDebug << " 帧:" << pInferenceResultData->iFrameId
                     << " 数据源:" << pInferenceResultData->iDataSource
                     << " --iClassId:" << it_result.iClassId
                     << " confidence=" << it_result.fScore
                     << " lx=" << it_result.fLTX
                     << " ly=" << it_result.fLTY
                     << " rx=" << it_result.fRBX
                     << " ry=" << it_result.fRBY;
        }

        outputQueMap_[strPort0_]->push(std::static_pointer_cast<void>(pInferenceResultData), true);
    }
    return APP_ERR_OK;
}

engine/SaveDebugImageEngine/SaveDebugImageEngine.cpp

@@ -43,6 +43,7 @@ APP_ERROR SaveDebugImageEngine::Process()
         }
 
         std::shared_ptr<VStep2OutputData> pVStep2OutputData = std::static_pointer_cast<VStep2OutputData>(pvoidd);
+        cv::Mat image = pVStep2OutputData->cvImage.clone();
 
         std::string strDataDir = this->baseConfig_.strDebugResultPath + "/"
                                  + pVStep2OutputData->strDetectDate + "/"
@@ -66,9 +67,9 @@ APP_ERROR SaveDebugImageEngine::Process()
 
 //        cv::Mat image = cv::imread(strImagePath);
 
-        if (pVStep2OutputData->cvImage.empty())
+        if (image.empty())
         {
-            LogWarn << " 帧:" << pVStep2OutputData->iFrameId << " 数据源:" << pVStep2OutputData->iDataSource << " debug图像未找到";
+//            LogWarn << " 帧:" << pVStep2OutputData->iFrameId << " 数据源:" << pVStep2OutputData->iDataSource << " debug图像未找到";
             continue;
         }
 
@@ -86,7 +87,7 @@ APP_ERROR SaveDebugImageEngine::Process()
                 cv::Size text_size = cv::getTextSize(i, cv::FONT_HERSHEY_SIMPLEX, 1, 2, 0);
                 cv::Point baseline_loc(text_org);
                 baseline_loc.y += base_line + text_size.height;
-                cv::putText(pVStep2OutputData->cvImage, i, baseline_loc, cv::FONT_HERSHEY_SIMPLEX, 1, cv::Scalar(0, 255, 0, 255), 2);
+                cv::putText(image, i, baseline_loc, cv::FONT_HERSHEY_SIMPLEX, 1, cv::Scalar(0, 255, 0, 255), 2);
                 // 手动调整x位置，为下一个单词留出空间
                 text_org.x += text_size.width + 10;
                 text_org.x = text_org.x > IMAGE_WIDTH ? 15 : text_org.x;
@@ -98,17 +99,17 @@ APP_ERROR SaveDebugImageEngine::Process()
             float centerX = pVStep2OutputData->step2ResultData.fLTX + (pVStep2OutputData->step2ResultData.fRBX - pVStep2OutputData->step2ResultData.fLTX)/2;
             float centerY = pVStep2OutputData->step2ResultData.fLTY + (pVStep2OutputData->step2ResultData.fRBY - pVStep2OutputData->step2ResultData.fLTY)/2;
 //        auto start = std::chrono::system_clock::now();  //计时开始
-            cv::rectangle(pVStep2OutputData->cvImage,
+            cv::rectangle(image,
                           cv::Point(pVStep2OutputData->step2ResultData.fLTX, pVStep2OutputData->step2ResultData.fLTY),
                           cv::Point(pVStep2OutputData->step2ResultData.fRBX, pVStep2OutputData->step2ResultData.fRBY),
                           cvScalar, 2);
-            cv::line(pVStep2OutputData->cvImage,
+            cv::line(image,
                      cv::Point(centerX, pVStep2OutputData->step2ResultData.fLTY-30), cv::Point(centerX, pVStep2OutputData->step2ResultData.fRBY+30),
                      cvScalar, 1);
 
             cv::Size text_size = cv::getTextSize(pVStep2OutputData->step2ResultData.transInfo.strTmpResult, cv::FONT_HERSHEY_SIMPLEX, 1, 2, 0);
             cv::Point linePoint(pVStep2OutputData->step2ResultData.fLTX, pVStep2OutputData->step2ResultData.fRBY + text_size.height + 5);
-            cv::putText(pVStep2OutputData->cvImage,
+            cv::putText(image,
                         pVStep2OutputData->step2ResultData.transInfo.strTmpResult,
                         linePoint,
                         cv::FONT_HERSHEY_SIMPLEX, 1, cv::Scalar(0, 255, 0, 180), 2);
@@ -118,11 +119,11 @@ APP_ERROR SaveDebugImageEngine::Process()
             {
                 cvScalar = {0, 255, 255, 255};
                 float centerX_corner = step2ResultData.fLTX + (step2ResultData.fRBX - step2ResultData.fLTX)/2;
-                cv::rectangle(pVStep2OutputData->cvImage,
+                cv::rectangle(image,
                               cv::Point(step2ResultData.fLTX, step2ResultData.fLTY),
                               cv::Point(step2ResultData.fRBX, step2ResultData.fRBY),
                               cvScalar, 2);
-                cv::line(pVStep2OutputData->cvImage,
+                cv::line(image,
                          cv::Point(centerX_corner, step2ResultData.fLTY-30), cv::Point(centerX_corner, step2ResultData.fRBY+30),
                          cvScalar, 1);
             }
@@ -131,7 +132,7 @@ APP_ERROR SaveDebugImageEngine::Process()
 //        auto end = std::chrono::system_clock::now();
 //        LogDebug << "图片存储用时: " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count() << "ms  " << strImagePath;
 
-        if (!cv::imwrite(strImagePath, pVStep2OutputData->cvImage, this->vecCompressionParams_))
+        if (!cv::imwrite(strImagePath, image, this->vecCompressionParams_))
         {
             LogError << "图片存储失败：" << strImagePath;
         }

engine/SaveDebugImageEngine/SaveDebugImageEngine.h

@@ -24,7 +24,7 @@ private:
     ai_matrix::BaseConfig baseConfig_;
 
     std::string strPort0_;
-    int iPicQuality_ = 100;
+    int iPicQuality_ = 80;
     std::vector<int> vecCompressionParams_;
 
 };

engine/SelectBestEngine/SelectBestEngine.cpp

@@ -99,10 +99,24 @@ void SelectBestEngine::selectBest()
     auto contains = [] (const std::vector<std::string>& vec, std::string value) -> bool {
         return std::find(vec.begin(), vec.end(), value) != vec.end();
     };
+    auto vCompare = [](std::string a, std::string b) {
+        if (a.size() != b.size()) return 999;
+
+        int count = 0;
+        for (int i = 0; i < a.size(); ++i)
+        {
+            if (a[i] != b[i])
+            {
+                ++count;
+            }
+        }
+        return count;
+    };
     DetectResultData detectResultData;
     detectResultData.strDetectDate = this->strDetectDate_;
     detectResultData.strDetectTime = this->strDetectTime_;
 
+    std::string strAlternativeResult;
     bool bHaveTwoContainer = false;
     for (auto & it_container : this->mapIndex_Containers_) {
         bool bHaveVerify = false;
@@ -129,21 +143,39 @@ void SelectBestEngine::selectBest()
                     count = it_max.second;
                     if (!contains(detectResultData.vecContainerNO, it_max.first)) {
                         strBestContainer = it_max.first;
+                        break;
+                    }
+                }
+            }
+            if (mapContainer_count.size() > 1)
+            {
+                for (auto &it_max : mapContainer_count)
+                {
+                    if (it_max.first != strBestContainer && vCompare(it_max.first, strBestContainer) > 2)
+                    {
+                        strAlternativeResult = it_max.first;
                     }
                 }
             }
             detectResultData.vecContainerNO.emplace_back(strBestContainer);
         } else {
-            // 取识别字数最长的
+            if (detectResultData.vecContainerNO.size() == 1 && !strAlternativeResult.empty())
-            std::string strBestContainer;
+            {
-            int iMaxSize = 0;
+                detectResultData.vecContainerNO.emplace_back(strAlternativeResult);
-            for (auto &it_max: mapContainer_count) {
+            }
-                if (it_max.first.size() > iMaxSize) {
+            else
-                    iMaxSize = it_max.first.size();
+            {
-                    strBestContainer = it_max.first;
+                // 取识别字数最长的
-                }
+                std::string strBestContainer;
+                int iMaxSize = 0;
+                for (auto &it_max: mapContainer_count) {
+                    if (it_max.first.size() > iMaxSize) {
+                        iMaxSize = it_max.first.size();
+                        strBestContainer = it_max.first;
+                    }
+                }
+                detectResultData.vecContainerNO.emplace_back(strBestContainer);
             }
-            detectResultData.vecContainerNO.emplace_back(strBestContainer);
         }
     }
     if (bHaveTwoContainer && detectResultData.vecContainerNO.size() < 2)
@@ -175,6 +207,12 @@ APP_ERROR SelectBestEngine::Process()
 
         if (pVSelectBestData->bIsEnd)
         {
+            if (this->strDetectDate_.empty())
+            {
+                this->strDetectDate_ = pVSelectBestData->strDetectDate;
+                this->strDetectTime_ = pVSelectBestData->strDetectTime;
+            }
+
             this->iEndCount_++;
             if (!(this->iEndCount_ % this->vecDataSourceConfig_.size()))
             {
@@ -183,6 +221,7 @@ APP_ERROR SelectBestEngine::Process()
             }
             continue;
         }
+        this->iEndCount_ = 0;
 
         if (this->strImagePath_.empty())
         {

engine/Step1InferenceEngine/ContainerStep1InferenceEngine.cpp

@@ -268,6 +268,7 @@ APP_ERROR ContainerStep1InferenceEngine::Process()
 
         if (pVDetectInfo->cvImage.empty())
         {
+            outputQueMap_[strPort0_]->push(std::static_pointer_cast<void>(pInferenceResultData), true);
             usleep(1000); //1ms
             continue;
         }
@@ -290,7 +291,7 @@ APP_ERROR ContainerStep1InferenceEngine::Process()
 
         pInferenceResultData->strDetectDate = pVDetectInfo->strDetectDate;
         pInferenceResultData->strDetectTime = pVDetectInfo->strDetectTime;
-        pInferenceResultData->cvImage = pVDetectInfo->cvImage;
+        pInferenceResultData->cvImage = pVDetectInfo->cvImage.clone();
 
         // 筛选离中心点坐标最近的，如果一样近就选最右的
         float fCenterX = IMAGE_WIDTH/2;
@@ -323,7 +324,7 @@ APP_ERROR ContainerStep1InferenceEngine::Process()
                 singledata.fLTY = inferenceResult.bbox[1];
                 singledata.fRBX = inferenceResult.bbox[2];
                 singledata.fRBY = inferenceResult.bbox[3];
-                singledata.fClear = inferenceResult.clear_id;
+                singledata.fClear = inferenceResult.clear_conf;
 
                 pInferenceResultData->singleData = singledata;