Home/ Questions/Q 9181999
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-17T18:27:21+00:00

I’m trying to locate orange balls from a video camera. I’m converting each frame

  • 0

I’m trying to locate orange balls from a video camera.

I’m converting each frame to HSV, blurring it, and using inRange to filter the orange color. I then apply the HoughCircles function but I can’t find the right parameters.

When it works, it’s not very stable: the circle is not always detected.

Here is a screenshot plus the code of the setup:

screenshot

#include <cv.h> 
#include <highgui.h> 
#include <stdio.h>  
#include <vector>

using namespace cv;

int hmin[2] = {2,0}, smin[2] = {163,104}, vmin[2] = {16,29};
int hmax[2] = {16,25}, smax[2] = {255, 255}, vmax[2] = {212,255};


int p1 = 6, p2 = 12;
int f1 = 9, f2 = 3;
int main() {
     VideoCapture cap(0);
     if(!cap.isOpened()) {
          fprintf( stderr, "ERROR: capture is NULL \n" );
          return -1;
     }

     namedWindow( "couleur", CV_WINDOW_AUTOSIZE );

     namedWindow( "filtre1", CV_WINDOW_AUTOSIZE );
     createTrackbar("hmin[0]", "filtre1", &hmin[0], 180, NULL);
     createTrackbar("smin[0]", "filtre1", &smin[0], 255, NULL);
     createTrackbar("vmin[0]", "filtre1", &vmin[0], 255, NULL);
     createTrackbar("hmax[0]", "filtre1", &hmax[0], 180, NULL);
     createTrackbar("smax[0]", "filtre1", &smax[0], 255, NULL);
     createTrackbar("vmax[0]", "filtre1", &vmax[0], 255, NULL);

     namedWindow( "hough", CV_WINDOW_AUTOSIZE );
     createTrackbar("p1", "hough", &p1, 500, NULL);
     createTrackbar("p2", "hough", &p2, 500, NULL);

     createTrackbar("f1", "hough", &f1, 30, NULL);
     createTrackbar("f2", "hough", &f2, 30, NULL);
     std::vector<Vec3f> cercle;

     while ( 1 ) {
          Mat frame, hsv, orange;
          cap >> frame;

          f1 = f1%2?f1:f1+1;

          cvtColor(frame, hsv, CV_BGR2HSV);
          GaussianBlur(hsv, hsv, Size(f1, f1), f2, f2 );
          inRange(hsv, Scalar(hmin[0],smin[0],vmin[0]), Scalar(hmax[0],smax[0],vmax[0]), orange);

          if(p1 <= 0)
               p1 = 1;

          if(p2 <= 0)
               p2 = 1;

          HoughCircles(orange, cercle, CV_HOUGH_GRADIENT, 2, orange.rows/4, p1, p2, 0, 30);

          for(uint i = 0; i < cercle.size(); i++) {
               Point c(cvRound(cercle[i][0]), cvRound(cercle[i][1]));
               int r = cvRound(cercle[i][2]);
               circle(frame, c, 3, Scalar(0,255,0), -1, 8, 0);
               circle(frame, c, r, Scalar(255,0,0), 3, 8, 0);
               printf("%d) %d %d %d\n", i, c.x, c.y, r);
          }
          printf("---\n");
          cercle.clear();

          imshow("couleur", frame );
          imshow("filtre1", orange );

          if ( (waitKey(10) & 255) == 27 ) break;
     }
     return 0;
}
Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    Here are some parameters that work.

    enter image description here enter image description here

    param1 (upper threshold for the internal Canny edge detector) is fixed at 100

    before passing image in to HoughCircles() the image is converted to grayscale and a median blur of ksize 5 is applied.

    Although this works for this one still frame, I am sure that a series of video frames will show a fair amount of variation and instability. For example, it’s quite tricky to find settings for this image that pick out both of the balls but don’t also select the white labels on the boxes.

    • 0