深度學習：駕駛行爲分析

一、功能與環境說明

程序功能簡介：使用yolo訓練，OpenCV調用、實現打哈欠、手機、抽菸、系安全帶，口罩檢測。

運行測試過的系統環境： 分別爲 windows系統、Linux系統、嵌入式Linux系統32位、嵌入式Linux系統64位。

運行測試過的硬件環境：分別爲普通筆記本電腦(i3、i5、i7)、RK3399、樹莓派4B

yolo環境搭建方法：https://pjreddie.com/darknet/yolo/

darknet框架安裝教程：https://pjreddie.com/darknet/install/

二、OpenCV調用代碼

2.1 .h頭文件代碼

#ifndef SDK_THREAD_H
#define SDK_THREAD_H
#include <QThread>
#include <QImage>

#include "opencv2/core/core.hpp"
#include "opencv2/core/core_c.h"
#include "opencv2/objdetect.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/imgproc.hpp"
#include <fstream>
#include <sstream>
#include <iostream>
#include <opencv2/dnn.hpp>
#include <opencv2/imgproc.hpp>
#include <vector>
#include <QDebug>

using namespace cv;
using namespace std;
using namespace dnn;

//視頻音頻編碼線程
class SDK_Thread: public QThread
{
    Q_OBJECT
public:
    void postprocess(Mat& frame, const vector<Mat>& outs, float confThreshold, float nmsThreshold);
    void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame);
    vector<String> getOutputsNames(Net&net);
    QImage Mat2QImage(const Mat& mat);
    Mat QImage2cvMat(QImage image);
protected:
    void run();
signals:
    void LogSend(QString text);
    void VideoDataOutput(QImage); //輸出信號
};
extern QImage save_image;
extern bool sdk_run_flag;
extern string names_file;
extern String model_def;
extern String weights;
#endif // SDK_THREAD_H

2.2 .cpp文件代碼

#include "sdk_thread.h"

QImage save_image; //用於行爲分析的圖片
bool sdk_run_flag=1;

Mat  SDK_Thread::QImage2cvMat(QImage image)
{
    Mat mat;
    switch(image.format())
    {
    case QImage::Format_ARGB32:
    case QImage::Format_RGB32:
    case QImage::Format_ARGB32_Premultiplied:
        mat = Mat(image.height(), image.width(), CV_8UC4, (void*)image.constBits(), image.bytesPerLine());
        break;
    case QImage::Format_RGB888:
        mat = Mat(image.height(), image.width(), CV_8UC3, (void*)image.constBits(), image.bytesPerLine());
        cvtColor(mat, mat, CV_BGR2RGB);
        break;
    case QImage::Format_Indexed8:
        mat = Mat(image.height(), image.width(), CV_8UC1, (void*)image.constBits(), image.bytesPerLine());
        break;
    }
    return mat;
}

QImage SDK_Thread::Mat2QImage(const Mat& mat)
{
    // 8-bits unsigned, NO. OF CHANNELS = 1
    if(mat.type() == CV_8UC1)
    {
        QImage image(mat.cols, mat.rows, QImage::Format_Indexed8);
        // Set the color table (used to translate colour indexes to qRgb values)
        image.setColorCount(256);
        for(int i = 0; i < 256; i++)
        {
            image.setColor(i, qRgb(i, i, i));
        }
        // Copy input Mat
        uchar *pSrc = mat.data;
        for(int row = 0; row < mat.rows; row ++)
        {
            uchar *pDest = image.scanLine(row);
            memcpy(pDest, pSrc, mat.cols);
            pSrc += mat.step;
        }
        return image;
    }
    // 8-bits unsigned, NO. OF CHANNELS = 3
    else if(mat.type() == CV_8UC3)
    {
        // Copy input Mat
        const uchar *pSrc = (const uchar*)mat.data;
        // Create QImage with same dimensions as input Mat
        QImage image(pSrc, mat.cols, mat.rows, mat.step, QImage::Format_RGB888);
        return image.rgbSwapped();
    }
    else if(mat.type() == CV_8UC4)
    {
        // Copy input Mat
        const uchar *pSrc = (const uchar*)mat.data;
        // Create QImage with same dimensions as input Mat
        QImage image(pSrc, mat.cols, mat.rows, mat.step, QImage::Format_ARGB32);
        return image.copy();
    }
    else
    {
        return QImage();
    }
}

vector<string> classes;
vector<String> SDK_Thread::getOutputsNames(Net&net)
{
    static vector<String> names;
    if (names.empty())
    {
        //Get the indices of the output layers, i.e. the layers with unconnected outputs
        vector<int> outLayers = net.getUnconnectedOutLayers();

        //get the names of all the layers in the network
        vector<String> layersNames = net.getLayerNames();

        // Get the names of the output layers in names
        names.resize(outLayers.size());
        for (size_t i = 0; i < outLayers.size(); ++i)
            names[i] = layersNames[outLayers[i] - 1];
    }
    return names;
}

void SDK_Thread::drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
{
    //Draw a rectangle displaying the bounding box
    rectangle(frame, Point(left, top), Point(right, bottom), Scalar(255, 178, 50), 3);

    //Get the label for the class name and its confidence
    string label = format("%.5f", conf);
    if (!classes.empty())
    {
        CV_Assert(classId < (int)classes.size());
        label = classes[classId] + ":" + label;
    }

    LogSend(tr("識別到的標籤:%1").arg(QString::fromStdString(label)));

    //Display the label at the top of the bounding box
    int baseLine;
    Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
    top = max(top, labelSize.height);
    rectangle(frame, Point(left, top - round(1.5*labelSize.height)), Point(left + round(1.5*labelSize.width), top + baseLine), Scalar(255, 255, 255), FILLED);
    putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.75, Scalar(0, 0, 0), 1);
}

void SDK_Thread::postprocess(Mat& frame, const vector<Mat>& outs, float confThreshold, float nmsThreshold)
{
    vector<int> classIds;
    vector<float> confidences;
    vector<Rect> boxes;

    for (size_t i = 0; i < outs.size(); ++i)
    {
        // Scan through all the bounding boxes output from the network and keep only the
        // ones with high confidence scores. Assign the box's class label as the class
        // with the highest score for the box.
        float* data = (float*)outs[i].data;
        for (int j = 0; j < outs[i].rows; ++j, data += outs[i].cols)
        {
            Mat scores = outs[i].row(j).colRange(5, outs[i].cols);
            Point classIdPoint;
            double confidence;
            // Get the value and location of the maximum score
            minMaxLoc(scores, 0, &confidence, 0, &classIdPoint);
            if (confidence > confThreshold)
            {
                int centerX = (int)(data[0] * frame.cols);
                int centerY = (int)(data[1] * frame.rows);
                int width = (int)(data[2] * frame.cols);
                int height = (int)(data[3] * frame.rows);
                int left = centerX - width / 2;
                int top = centerY - height / 2;

                classIds.push_back(classIdPoint.x);
                confidences.push_back((float)confidence);
                boxes.push_back(Rect(left, top, width, height));
            }
        }
    }

    // Perform non maximum suppression to eliminate redundant overlapping boxes with
    // lower confidences
    vector<int> indices;
    NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
    for (size_t i = 0; i < indices.size(); ++i)
    {
        int idx = indices[i];
        Rect box = boxes[idx];
        drawPred(classIds[idx], confidences[idx], box.x, box.y,
            box.x + box.width, box.y + box.height, frame);
    }
}

string names_file;
String model_def;
String weights;

void SDK_Thread::run()
{
    Mat frame,frame_src,blob;
    int inpWidth, inpHeight;
    //String names_file = "D:/linux-share-dir/yolo_v3/car/yolo.names";
    //String model_def = "D:/linux-share-dir/yolo_v3/car/yolov3.cfg";
    //String weights = "D:/linux-share-dir/yolo_v3/car/yolov3.weights";

    double thresh = 0.5;
    double nms_thresh = 0.4; //0.4  0.25
    inpWidth = inpHeight = 320; //416 608

    //read names
    ifstream ifs(names_file.c_str());
    string line;
    while(getline(ifs, line))classes.push_back(line);

    //初始化模型
    Net net = readNetFromDarknet(model_def, weights);
    net.setPreferableBackend(DNN_BACKEND_OPENCV);
    net.setPreferableTarget(DNN_TARGET_CPU);

    QImage use_image;
    LogSend("開始進行行爲分析.\n");

    //VideoCapture capture(0);// VideoCapture:OENCV中新增的類，捕獲視頻並顯示出來

    while(sdk_run_flag)
    {
       if(save_image.isNull())continue;
       LogSend(tr("開始識別------>\n"));
        //capture >> frame;
        //得到源圖片
        //use_image.load("D:/linux-share-dir/car_1.jpg");
        frame_src=QImage2cvMat(save_image);
        cvtColor(frame_src, frame, CV_RGB2BGR);
        //frame=imread("D:/linux-share-dir/car_1.jpg");//
        //frame=imread("D:/linux-share-dir/5.jpg");
        //in_w=save_image.width();
        //in_h=save_image.height();
        blobFromImage(frame, blob, 1/255.0, cvSize(inpWidth, inpHeight), Scalar(0,0,0), true, false);

        vector<Mat> mat_blob;
        imagesFromBlob(blob, mat_blob);

        //Sets the input to the network
        net.setInput(blob);

        // 運行前向傳遞以獲取輸出層的輸出
        vector<Mat> outs;
        net.forward(outs, getOutputsNames(net));

        postprocess(frame, outs, thresh, nms_thresh);

        vector<double> layersTimes;
        double freq = getTickFrequency() / 1000;
        double t = net.getPerfProfile(layersTimes) / freq;
       // string label = format("time : %.2f ms", t);
        //putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 0, 255));
        LogSend(tr("識別結束---消耗的時間:%1 秒\n").arg(t/1000));
        //得到處理後的圖像
        use_image=Mat2QImage(frame);
        use_image=use_image.rgbSwapped();
        VideoDataOutput(use_image.copy());
    }
}