光流法代码

#include "opencv2/opencv.hpp"
#include <sstream>
#include <iomanip>
using namespace std;
using namespace cv;

class FrameProcessor;
//帧处理基类
class FrameProcessor{
    public:
        virtual void process(Mat &input,Mat &ouput)=0;
};

//特征跟踪类，继承自帧处理基类
class FeatureTracker :  public FrameProcessor{
    Mat gray;  //当前灰度图
    Mat gray_prev;  //之前的灰度图
    vector<Point2f> points[2];//前后两帧的特征点
    vector<Point2f> initial;//初始特征点
    vector<Point2f> features;//检测到的特征
    int max_count; //要跟踪特征的最大数目
    double qlevel; //特征检测的指标
    double minDist;//特征点之间最小容忍距离
    vector<uchar> status; //特征跟踪状态
    vector<float> err; //跟踪时的错误
public:
    FeatureTracker():max_count(500),qlevel(0.01),minDist(10.){}
    void process(Mat &frame,Mat &output){
        //得到灰度图
        cvtColor (frame,gray,CV_BGR2GRAY);
        frame.copyTo (output);
        //特征点太少了，重新检测特征点
        if(addNewPoint()){
            detectFeaturePoint ();
            //插入检测到的特征点
            points[0].insert (points[0].end (),features.begin (),features.end ());
            initial.insert (initial.end (),features.begin (),features.end ());
        }
        //第一帧
        if(gray_prev.empty ()){
                gray.copyTo (gray_prev);
        }
        //根据前后两帧灰度图估计前一帧特征点在当前帧的位置
        //默认窗口是15*15
        calcOpticalFlowPyrLK (
                gray_prev,//前一帧灰度图
                gray,//当前帧灰度图
                points[0],//前一帧特征点位置
                points[1],//当前帧特征点位置
                status,//特征点被成功跟踪的标志
                err);//前一帧特征点点小区域和当前特征点小区域间的差，根据差的大小可删除那些运动变化剧烈的点
        int k = 0;
        //去除那些未移动的特征点
        for(int i=0;i<points[1].size ();i++){
            if(acceptTrackedPoint (i)){
                initial[k]=initial[i];
                points[1][k++] = points[1][i];
            }
        }
        points[1].resize (k);
        initial.resize (k);
        //标记被跟踪的特征点
        handleTrackedPoint (frame,output);
        //为下一帧跟踪初始化特征点集和灰度图像
        std::swap(points[1],points[0]);
        cv::swap(gray_prev,gray);
    }

    void detectFeaturePoint(){
        goodFeaturesToTrack (gray,//图片
                                 features,//输出特征点
                                 max_count,//特征点最大数目
                                 qlevel,//质量指标
                                 minDist);//最小容忍距离
    }
    bool addNewPoint(){
            //若特征点数目少于10，则决定添加特征点
        return points[0].size ()<=10;
    }

    //若特征点在前后两帧移动了，则认为该点是目标点，且可被跟踪
    bool acceptTrackedPoint(int i){
        return status[i]&&
                (abs(points[0][i].x-points[1][i].x)+
                  abs(points[0][i].y-points[1][i].y) >2);
    }

    //画特征点
    void  handleTrackedPoint(Mat &frame,Mat &output){
            for(int i=0;i<points[i].size ();i++){
                //当前特征点到初始位置用直线表示
                line(output,initial[i],points[1][i],Scalar::all (0));
                //当前位置用圈标出
                circle(output,points[1][i],3,Scalar::all(0),(-1));
            }
        }
};

class VideoProcessor{
private:
    VideoCapture caputure;
    //写视频流对象
    VideoWriter writer;
    //输出文件名
    string Outputfile;

    int currentIndex;
    int digits;
    string extension;
    FrameProcessor *frameprocessor;
    //图像处理函数指针
    void (*process)(Mat &,Mat &);
    bool callIt;
    string WindowNameInput;
    string WindowNameOutput;
    //延时
    int delay;
    long fnumber;
    //第frameToStop停止
    long frameToStop;
    //暂停标志
    bool stop;
    //图像序列作为输入视频流
    vector<string> images;
    //迭代器
public:
    VideoProcessor() : callIt(true),delay(0),fnumber(0),stop(false),digits(0),frameToStop(-1){}
   //设置图像处理函数
    void setFrameProcessor(void (*process)(Mat &,Mat &)){
        frameprocessor = 0;
        this->process = process;
        CallProcess ();
    }
    //打开视频
    bool setInput(string filename){
        fnumber = 0;
        //若已打开，释放重新打开
        caputure.release ();
        return caputure.open (filename);
    }
    //设置输入视频播放窗口
    void displayInput(string wn){
        WindowNameInput = wn;
        namedWindow (WindowNameInput);
    }
    //设置输出视频播放窗口
    void displayOutput(string wn){
        WindowNameOutput = wn;
        namedWindow (WindowNameOutput);
    }
    //销毁窗口
    void dontDisplay(){
        destroyWindow (WindowNameInput);
        destroyWindow (WindowNameOutput);
        WindowNameInput.clear ();
        WindowNameOutput.clear ();
    }

    //启动
    void run(){
        Mat frame;
        Mat output;
        if(!isOpened())
            return;
        stop = false;
        while(!isStopped()){
            //读取下一帧
            if(!readNextFrame(frame))
                break;
            if(WindowNameInput.length ()!=0)
                imshow (WindowNameInput,frame);
            //处理该帧
            if(callIt){
                if(process)
                    process(frame,output);
                else if(frameprocessor)
                    frameprocessor->process (frame,output);
            }
            else{
                output = frame;
            }
            if(Outputfile.length ()){
                    cvtColor (output,output,CV_GRAY2BGR);
                    writeNextFrame (output);
              }
            if(WindowNameOutput.length ()!=0)
                imshow (WindowNameOutput,output);
            //按键暂停，继续按键继续
            if(delay>=0&&waitKey (delay)>=0)
                waitKey(0);
            //到达指定暂停键，退出
            if(frameToStop>=0&&getFrameNumber()==frameToStop)
                stopIt();
        }
    }
    //暂停键置位
    void stopIt(){
        stop = true;
    }
    //查询暂停标志位
    bool isStopped(){
        return stop;
    }
    //返回视频打开标志
    bool isOpened(){
       return  caputure.isOpened ()||!images.empty ();
    }
    //设置延时
    void setDelay(int d){
        delay = d;
    }
    //读取下一帧
    bool readNextFrame(Mat &frame){
        if(images.size ()==0)
            return caputure.read (frame);
        else{
            if(itImg!=images.end()){
                frame = imread (*itImg);
                itImg++;
                return frame.data?1:0;
            }
            else
                return false;
        }
    }

    void CallProcess(){
        callIt = true;
    }
    void  dontCallProcess(){
        callIt = false;
    }
    //设置停止帧
    void stopAtFrameNo(long frame){
        frameToStop = frame;
    }
   // 获得当前帧的位置
    long getFrameNumber(){
        long fnumber = static_cast<long>(caputure.get ((CV_CAP_PROP_POS_FRAMES)));
        return fnumber;
    }

     //获得帧大小
       Size getFrameSize() {
        if (images.size()==0) {
            // 从视频流获得帧大小
            int w= static_cast<int>(caputure.get(CV_CAP_PROP_FRAME_WIDTH));
            int h= static_cast<int>(caputure.get(CV_CAP_PROP_FRAME_HEIGHT));
            return Size(w,h);
            }
        else {
                //从图像获得帧大小
                cv::Mat tmp= cv::imread(images[0]);
                return (tmp.data)?(tmp.size()):(Size(0,0));
            }
          }

   //获取帧率
    double getFrameRate(){
        return caputure.get(CV_CAP_PROP_FPS);
    }
    vector<string>::const_iterator itImg;
    bool setInput (const vector<string> &imgs){
        fnumber = 0;
        caputure.release ();
        images = imgs;
        itImg = images.begin ();
        return true;
    }

    void  setFrameProcessor(FrameProcessor *frameprocessor){
        process = 0;
        this->frameprocessor = frameprocessor;
        CallProcess ();
    }

    //获得编码类型
    int getCodec(char codec[4]) {
        if (images.size()!=0)
            return -1;
        union { // 数据结构4-char
            int value;
            char code[4];
        } returned;
        //获得编码值
        returned.value= static_cast<int>(
        caputure.get(CV_CAP_PROP_FOURCC));
        // get the 4 characters
        codec[0]= returned.code[0];
        codec[1]= returned.code[1];
        codec[2]= returned.code[2];
        codec[3]= returned.code[3];
        return returned.value;
    }

    bool setOutput(const string &filename,int codec = 0,double framerate = 0.0,bool isColor = true){
        //设置文件名
        Outputfile = filename;
        //清空扩展名
        extension.clear ();
        //设置帧率
        if(framerate ==0.0){
            framerate = getFrameRate ();
        }
        //获取输入原视频的编码方式
        char c[4];
        if(codec==0){
            codec = getCodec(c);
        }
        return writer.open(Outputfile,
                           codec,
                           framerate,
                           getFrameSize(),
                           isColor);
    }

    //输出视频帧到图片fileme+currentIndex.ext,如filename001.jpg
    bool setOutput (const string &filename,//路径
                    const string &ext,//扩展名
                    int numberOfDigits=3,//数字位数
                    int startIndex=0 ){//起始索引
           if(numberOfDigits<0)
               return false;
           Outputfile = filename;
           extension = ext;
           digits = numberOfDigits;
           currentIndex = startIndex;
           return true;
    }

    //写下一帧
    void writeNextFrame(Mat &frame){
        //如果扩展名不为空，写到图片文件中
        if(extension.length ()){
            stringstream ss;
            ss<<Outputfile<<setfill('0')<<setw(digits)<<currentIndex++<<extension;
            imwrite (ss.str (),frame);
        }
        //反之，写到视频文件中
        else{
            writer.write (frame);
        }
    }

};

    //帧处理函数：canny边缘检测
    void canny(cv::Mat& img, cv::Mat& out) {
        //灰度变换
        if (img.channels()==3)
            cvtColor(img,out,CV_BGR2GRAY);
        // canny算子求边缘
        Canny(out,out,100,200);
        //颜色反转，看起来更舒服些
        threshold(out,out,128,255,cv::THRESH_BINARY_INV);
    }

int main(int argc, char *argv[])
{
    VideoProcessor processor;
    FeatureTracker tracker;
    //打开输入视频
    processor.setInput ("bike.avi");
    processor.displayInput ("Current Frame");
    processor.displayOutput ("Output Frame");
    //设置每一帧的延时
    processor.setDelay (1000./processor.getFrameRate ());
    //设置帧处理函数，可以任意
    processor.setFrameProcessor (&tracker);
    //   processor.setOutput ("./bikeout.avi");
    //    processor.setOutput ("bikeout",".jpg");
    processor.run ();
    return 0;
}
摘自：