find circles

approxPolyDP

#include <opencv2/opencv.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
using namespace cv;
using namespace std;
#define  _ShowImage 1
////////////////////////////////////////////////////////
/**
* @name  getAllImgsFilePath
* @brief 获取指定路径下的所有文件全路径(还是这个方便啊.)
* @param[in] szDir:指定路径(比如C:\\_Sunyard\\Project\\浙江农信\\data\\照片\\Merge表格\\*.*";)
* @param[in/out]  testImgFiles,所有文件全路径存在这里
* @return 0 成功 / -1 失败
* @author [lintao 2015-10-09]
**/
////////////////////////////////////////////////////////
int getAllImgsFilePath(char *szDir,vector<string>& testImgFiles)
{
    WIN32_FIND_DATA ffd;
    HANDLE hFind = FindFirstFile((szDir), &ffd);
    //testImgFiles.push_back(ffd.cFileName);//do 是先做..会push进去的
    // List all the files in the directory with some info about them.
    do
    {
        if (ffd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
        {
            //_tprintf(TEXT("  %s   <DIR>\n"), ffd.cFileName);
            //cout<<ffd.cFileName<<endl;
        }
        else
        {
            //_tprintf(TEXT("  %s  \n"), ffd.cFileName);
            //cout<<ffd.cFileName<<endl;
            string searchImgDirWithType =szDir;                         //搜索目录带有搜索文件类型的
            int npos=searchImgDirWithType.find_last_of("\\");           //
            string searchImgDir =searchImgDirWithType.substr(0,npos+1); //纯目录    //substr 返回pos开始的n个字符组成的字符串
            string imgWholePath = searchImgDir + ffd.cFileName;
            testImgFiles.push_back(imgWholePath);
        }
    } while (FindNextFile(hFind, &ffd) != 0);
    FindClose(hFind);
    return 0;
}
//////////////////////////////////////////////////////////////////////////
//判断轮廓是不是一个圆，通过轮廓的面积pi*r*r的差异，判断是不是圆。
//contour：输入轮廓
//minRadius: 半径长度至少大于minRadius
//////////////////////////////////////////////////////////////////////////
BOOL isCircleContour(vector<Point>& contour,double minRadius)
{
    BOOL isCircle = FALSE;
    //double perimeter = arcLength(contour,TRUE);
    double perimeter = arcLength(contour,FALSE);
    double perimeter2 =    arcLength(contour,TRUE);
    Rect rect =boundingRect(contour);
    double r = (rect.width+rect.height)/4.0;    //半径约等于轮廓外接矩的边长/2
    double area=contourArea(contour);
    double pi=3.1415926;
    //cout<<"x/y/width/height\t"<<rect.x<<"  "<<rect.y<<"  "<<rect.width<<"  "<<rect.height<<endl;
    //cout<<"perimeter:FALSE\t"<<perimeter<<endl;
    //cout<<"perimeter2:TRUE\t"<<perimeter2<<endl;
    //cout<<"area:\t"<<area<<endl;
    //cout<<"r: "<<r<<"  计算出周长: "<<2*pi*r<<"  计算出面积: "<<pi*r*r<<endl;
    double perimeterDiffRatio =    fabs(2*pi*r    - perimeter)*2.0/(2*pi*r + perimeter) *100;
    double areaDiffRatio = fabs(pi*r*r - area)*2.0/(pi*r*r + area) *100;
    double sideDiffRatio = abs(rect.width-rect.height)*2.0/(rect.width+rect.height)*100;
    //CString CstrInfo3;
    //CstrInfo3.Format(_T(" perimeterDiffRatio: %f\n areaDiffRatio: %f \n  sideDiffRatio:%f"),perimeterDiffRatio,areaDiffRatio,sideDiffRatio);
    //AfxMessageBox(CstrInfo3);
    if (r>minRadius)
    {
        cout<<"周长差异百分比:  "<<perimeterDiffRatio<<"%"<<endl;;
        cout<<"面积差异百分比:  "<<areaDiffRatio<<"%"<<endl;
        cout<<"边长差异百分比:  "<<sideDiffRatio<<"%"<<endl;;
        cout<<endl<<endl;
    }
    //原始的
    //const double areaDiffThresh = 18.0;       //目前统计的，最大的面积差2.58%
    //const double perimeterDiffThresh = 15;    //目前统计的，最大的周长差5.65%
    //const double sideDiffThresh = 15;     //目前统计的，最大的边长差7.3%
    //const double avgDiffThresh = 15.0;        //三个差异比值的加权平均要求小于5.0;
    const double areaDiffThresh = 5;        //目前统计的，最大的面积差2.58%
    const double perimeterDiffThresh =7.25; //目前统计的，最大的周长差5.65%
    const double sideDiffThresh = 10;       //目前统计的，最大的边长差7.3%
    const double avgDiffThresh = 5.0;       //三个差异比值的加权平均要求小于5.0;
    double avdDiffRatio = (areaDiffRatio+perimeterDiffRatio+sideDiffRatio)/3.0;
    //计算出的轮廓周长是半径的2*pi倍数
    if( (areaDiffRatio <areaDiffThresh)
        && (perimeterDiffRatio <perimeterDiffThresh)
        && (sideDiffRatio <sideDiffThresh)
        &&(avdDiffRatio<avgDiffThresh)
        && (r>minRadius) )
    {
        isCircle = TRUE;
    }
    return isCircle;
}
int countShootingRingsNum(Mat& orgImg)
{
    if (orgImg.empty())
        return -1;
    //转换到灰度图
    Mat grayImage;
    if (orgImg.channels()==3)
        cvtColor(orgImg,grayImage,CV_BGR2GRAY);
    if (orgImg.channels()==1)
        grayImage = orgImg.clone();
    imshow("grayImage",grayImage);
    //模块：二值化
    int isBlackCharacter= FALSE;
    Mat binImg;
    double otsuThresh = 0.0;
    if(isBlackCharacter)
        otsuThresh = threshold(grayImage, binImg,  0, 255, CV_THRESH_BINARY_INV|CV_THRESH_OTSU);
    else
        otsuThresh = threshold(grayImage, binImg,  0, 255, CV_THRESH_BINARY|CV_THRESH_OTSU);
    //CString CstrInfo3;
    //CstrInfo3.Format(_T(" otsuThresh: %f"),otsuThresh);
    //AfxMessageBox(CstrInfo3);
    imshow("binImg",binImg);
    //腐蚀膨胀
    int erodeSize = 3;
    int dilateSize = 3;
    Mat elementForErode = getStructuringElement(MORPH_RECT, Size(erodeSize, erodeSize), Point(-1, -1) );
    Mat elementForDilate = getStructuringElement(MORPH_RECT, Size(dilateSize, dilateSize), Point(-1, -1) );
    dilate(binImg ,binImg , elementForDilate);
    //erode(binImg ,binImg , elementForErode );
    imshow("膨胀腐蚀后的Binary Image", binImg);
    //waitKey();
    //模块：提取轮廓(保留轮廓面积大于_tooLessPixels)
    vector< vector<Point> > contours;   // 轮廓
    vector< vector<Point> > filterContours; // 筛选后的轮廓
    vector< Vec4i > hierarchy;    // 轮廓的结构信息
    contours.clear();
    hierarchy.clear();
    filterContours.clear();
    Mat binaryImageCopy1 = binImg.clone();
    //findContours(binaryImageCopy1, contours, hierarchy, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
    findContours(binaryImageCopy1, contours, hierarchy, CV_RETR_CCOMP, CV_CHAIN_APPROX_SIMPLE);
    // 去除伪轮廓: 轮廓上像素点数太少的
    const int _tooLessPixels = 10*10;
    for (size_t i = 0; i < contours.size(); i++)
    {
        if (fabs(contourArea(contours[i])) > _tooLessPixels)
        {
            //先都push进去
            filterContours.push_back(contours[i]);
        }
    }
    Mat rgbImg = binImg.clone();
    cvtColor(rgbImg,rgbImg,CV_GRAY2BGR);
    //binaryImage.setTo(0);
    drawContours(rgbImg, filterContours, -1, Scalar(0,255,0), 5); //8, hierarchy);
    //cvNamedWindow("filterContours",0);
    imshow("filterContours",rgbImg);
    //模块：外接矩的宽高比比例失调的过滤掉
    const double w2hThresh2 = 2.0;      //外接矩的宽高比比例大于3过滤掉
    const double h2wThresh2 = 2.0;      //外接矩的宽高比比例大于3过滤掉
    for (int i=0; i<filterContours.size(); i++)
    {
        Rect tmpRect = boundingRect(filterContours[i]);
        if( //(tmpRect.width>maxRectWidth)
            // (tmpRect.height>maxRectHeight)
            ( (double)tmpRect.width/(double)tmpRect.height > w2hThresh2)
            || ( (double)tmpRect.height/(double)tmpRect.width > h2wThresh2)
            )
        {
            filterContours.erase(filterContours.begin()+i);
            i--;//删除掉一个后，索引回退一个
        }
    }
    //不是圆的过滤掉
    double minRadius = 20/2;            //目前最小的圆直径30
    for (int i=0; i<filterContours.size(); i++)
    {
        if(!isCircleContour(filterContours[i], minRadius))
        {
            //selectContours.push_back(filterContours[i]);
            filterContours.erase(filterContours.begin()+i);
            i--;//删除掉一个后，索引回退一个
        }
    }
    Mat rgbImg2 = binImg.clone();
    cvtColor(rgbImg2,rgbImg2,CV_GRAY2BGR);
    //binaryImage.setTo(0);
    drawContours(rgbImg2, filterContours, -1, Scalar(0,255,0), 5); //8, hierarchy);
    //cvNamedWindow("filterContours2",0);
    imshow("filterContours2",rgbImg2);
    Mat grayImg;
    cvtColor(orgImg,grayImg,CV_BGR2GRAY);
    Mat diffImg;
    absdiff(grayImg,GreenImg,diffImg);
    normalize(diffImg, diffImg, 0, 255, NORM_MINMAX);
    imshow("绿色通道和灰度图差异图图",diffImg);
    return 0;
}
int testCountShootingRingsNum()
{
    //char *szDir = "C:\\Projects\\ComputerVisionProject\\网上看到的项目\\数打靶的环数\\data\\webData\\*.*";
    char *szDir = "C:\\Projects\\ComputerVisionProject\\网上看到的项目\\数打靶的环数\\data\\*.*";
    vector<string> testImgFiles;
    getAllImgsFilePath(szDir,testImgFiles);
    //system("time");
    for (int i=0; i<testImgFiles.size(); i++)
    {
        //输出文件名
        int npos=testImgFiles[i].find_last_of("\\");
        string fileName =testImgFiles[i].substr(npos+1,testImgFiles[i].length()-(npos+1));  //获取目录
        cout<<fileName<<endl;
        Mat orgImg =imread(testImgFiles[i]);
        if (orgImg.empty())
            continue;
        imshow("orgImg",orgImg);
        countShootingRingsNum(orgImg);
        waitKey();
    }
    system("time");
    return 0;
}
void main()
{
    testCountShootingRingsNum();
}