OpenCV detect partial circle with noise

using this as input (your own median filtered image (I've just cropped it):

enter image description here

First I "normalize" the image. I just stretch values, that smallest val is 0 and biggest val is 255, leading to this result: (maybe some real contrast enhancement is better)

enter image description here

after that I compute the threshold of that image with some fixed threshold (you might need to edit that and find a way to choose the threshold dynamically! a better contrast enhancement might help there)

enter image description here

from this image, I use some simple RANSAC circle detection(very similar to my answer in the linked semi-circle detection question), giving you this result as a best semi-sircle:

enter image description here

int main()


    //cv::Mat color = cv::imread("../inputData/semi_circle_contrast.png");

    cv::Mat color = cv::imread("../inputData/semi_circle_median.png");

    cv::Mat gray;


    // convert to grayscale

    cv::cvtColor(color, gray, CV_BGR2GRAY);


    // now map brightest pixel to 255 and smalles pixel val to 0. this is for easier finding of threshold

    double min, max;


    float sub = min;

    float mult = 255.0f/(float)(max-sub);

    cv::Mat normalized = gray - sub;

    normalized = mult * normalized;

    cv::imshow("normalized" , normalized);




    // now compute threshold

    // TODO: this might ne a tricky task if noise differs...

    cv::Mat mask;

    //cv::threshold(input, mask, 0, 255, CV_THRESH_BINARY | CV_THRESH_OTSU);

    cv::threshold(normalized, mask, 100, 255, CV_THRESH_BINARY);




    std::vector<cv::Point2f> edgePositions;

    edgePositions = getPointPositions(mask);


    // create distance transform to efficiently evaluate distance to nearest edge

    cv::Mat dt;

    cv::distanceTransform(255-mask, dt,CV_DIST_L1, 3);


    //TODO: maybe seed random variable for real random numbers.


    unsigned int nIterations = 0;


    cv::Point2f bestCircleCenter;

    float bestCircleRadius;

    float bestCirclePercentage = 0;

    float minRadius = 50;   // TODO: ADJUST THIS PARAMETER TO YOUR NEEDS, otherwise smaller circles wont be detected or "small noise circles" will have a high percentage of completion


    //float minCirclePercentage = 0.2f;

    float minCirclePercentage = 0.05f;  // at least 5% of a circle must be present? maybe more...


    int maxNrOfIterations = edgePositions.size();   // TODO: adjust this parameter or include some real ransac criteria with inlier/outlier percentages to decide when to stop


    for(unsigned int its=0; its< maxNrOfIterations; ++its)


        //RANSAC: randomly choose 3 point and create a circle:

        //TODO: choose randomly but more intelligent,

        //so that it is more likely to choose three points of a circle.

        //For example if there are many small circles, it is unlikely to randomly choose 3 points of the same circle.

        unsigned int idx1 = rand()%edgePositions.size();

        unsigned int idx2 = rand()%edgePositions.size();

        unsigned int idx3 = rand()%edgePositions.size();


        // we need 3 different samples:

        if(idx1 == idx2) continue;

        if(idx1 == idx3) continue;

        if(idx3 == idx2) continue;


        // create circle from 3 points:

        cv::Point2f center; float radius;



        // inlier set unused at the moment but could be used to approximate a (more robust) circle from alle inlier

        std::vector<cv::Point2f> inlierSet;


        //verify or falsify the circle by inlier counting:

        float cPerc = verifyCircle(dt,center,radius, inlierSet);


        // update best circle information if necessary

        if(cPerc >= bestCirclePercentage)

            if(radius >= minRadius)


            bestCirclePercentage = cPerc;

            bestCircleRadius = radius;

            bestCircleCenter = center;





    // draw if good circle was found

    if(bestCirclePercentage >= minCirclePercentage)

        if(bestCircleRadius >= minRadius);

        cv::circle(color, bestCircleCenter,bestCircleRadius, cv::Scalar(255,255,0),1);







        return 0;



float verifyCircle(cv::Mat dt, cv::Point2f center, float radius, std::vector<cv::Point2f> & inlierSet)
 unsigned int counter = 0;
 unsigned int inlier = 0;
 float minInlierDist = 2.0f;
 float maxInlierDistMax = 100.0f;
 float maxInlierDist = radius/25.0f;
 if(maxInlierDist<minInlierDist) maxInlierDist = minInlierDist;
 if(maxInlierDist>maxInlierDistMax) maxInlierDist = maxInlierDistMax;
 // choose samples along the circle and count inlier percentage
 for(float t =0; t<2*3.14159265359f; t+= 0.05f)
     float cX = radius*cos(t) + center.x;
     float cY = radius*sin(t) + center.y;
     if(cX < dt.cols)
     if(cX >= 0)
     if(cY < dt.rows)
     if(cY >= 0)
     if(<float>(cY,cX) < maxInlierDist)
 return (float)inlier/float(counter);
inline void getCircle(cv::Point2f& p1,cv::Point2f& p2,cv::Point2f& p3, cv::Point2f& center, float& radius)
  float x1 = p1.x;
  float x2 = p2.x;
  float x3 = p3.x;
  float y1 = p1.y;
  float y2 = p2.y;
  float y3 = p3.y;
  center.x = (x1*x1+y1*y1)*(y2-y3) + (x2*x2+y2*y2)*(y3-y1) + (x3*x3+y3*y3)*(y1-y2);
  center.x /= ( 2*(x1*(y2-y3) - y1*(x2-x3) + x2*y3 - x3*y2) );
  center.y = (x1*x1 + y1*y1)*(x3-x2) + (x2*x2+y2*y2)*(x1-x3) + (x3*x3 + y3*y3)*(x2-x1);
  center.y /= ( 2*(x1*(y2-y3) - y1*(x2-x3) + x2*y3 - x3*y2) );
  radius = sqrt((center.x-x1)*(center.x-x1) + (center.y-y1)*(center.y-y1));
std::vector<cv::Point2f> getPointPositions(cv::Mat binaryImage)
 std::vector<cv::Point2f> pointPositions;
 for(unsigned int y=0; y<binaryImage.rows; ++y)
     //unsigned char* rowPtr = binaryImage.ptr<unsigned char>(y);
     for(unsigned int x=0; x<binaryImage.cols; ++x)
         //if(rowPtr[x] > 0) pointPositions.push_back(cv::Point2i(x,y));
         if(<unsigned char>(y,x) > 0) pointPositions.push_back(cv::Point2f(x,y));
 return pointPositions;


posted on 2017-10-17 13:39 zmj 阅读(603) 评论(0)  编辑 收藏 引用

【推荐】超50万行VC++源码: 大型组态工控、电力仿真CAD与GIS源码库
网站导航: 博客园   IT新闻   BlogJava   知识库   博问   管理