Revision history [back]

Hi, I need to improve my plate poligon on a low res image (80x20 dots).

The first step (classifier) works perfect. Then I run the following script to find the polygons to apply perspective correction(also works great). The only problem the detection works only for few plates.

My question is how to improve it (tunning the parameters or changing the detection method).

---Find contours and approx a Poly..........

cv::findContours(toPrespective, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);

        for (int j = 0; j < contours.size(); j++)
        {

            cv::approxPolyDP(cv::Mat(contours[j]), approx, cv::arcLength(cv::Mat(contours[j]), true)*0.01, true);

            if (std::fabs(cv::contourArea(contours[j])) < 150 || !cv::isContourConvex(approx))
                continue;

            if (approx.size() == 4)
            {


                  .... And then the deskew stuff that works prefect.......

THANKS