c++ - Proper way to determine thresholding parameters -

i trying find triangles (blue contours) , trapezoids (yellow contours) in real time. in general it's okay.
there problems. first it's false positives. triangles become trapezoids , vice versa. , don't know how how solve problem. second it's "noise". . tried check area of figure, noise can equal area. did not much. noise depends on thresholding parameters. cv::adaptivethresholddoes not @ all. it's adds more noise (and slow) erode , dilate cant fix in proper way

and here code.

cv::mat detect(cv::mat imagergb) {     //rgb -> gray     cv::mat imagegray;     cv::cvtcolor(imagergb, imagegray, cv_bgr2gray);     //bluring     cv::mat image;     cv::gaussianblur(imagegray, image, cv::size(5,5), 2);     //thresholding     cv::threshold(image, image, 100, 255, cv_thresh_binary_inv);      //slow , noise     //cv::adaptivethreshold(image, image, 255.0, cv_adaptive_thresh_gaussian_c, cv_thresh_binary, 21, 0);      //calculating canny params.     cv::scalar mu;     cv::scalar sigma;     cv::meanstddev(image, mu, sigma);      cv::mat imagecanny;      cv::canny(image,               imagecanny,               mu.val[0] + sigma.val[0],               mu.val[0] - sigma.val[0]);      //detecting conturs.     std::vector<std::vector<cv::point> > contours;     std::vector<cv::vec4i> hierarchy;     cv::findcontours(imagecanny, contours, hierarchy,cv_retr_tree, cv_chain_approx_none);      //hierarchy not needed here clear it.     hierarchy.clear();      (std::size_t = 0; < contours.size(); i++)     {         //fitellipse need @ last 5 points.         if (contours.at(i).size() < 5)         {             continue;         }         //skip small contours.         if (std::fabs(cv::contourarea(contours.at(i))) < 800.0)         {             continue;         }         //calculating rotatedrect contours not hull         //because fitellipse need @ last 5 points.          cv::rotatedrect bellipse = cv::fitellipse(contours.at(i));          //finds convex hull of point set.         std::vector<cv::point> hull;         cv::convexhull(contours.at(i), hull, true);         //approx it, we'll 3 point triangles         //and 4 points trapez.         cv::approxpolydp(hull, hull, 15, true);         //is our contour convex. it's mast be.         if (!cv::iscontourconvex(hull))         {             continue;         }         //triangle         if (hull.size() == 3)         {             cv::drawcontours(imagergb, contours, i, cv::scalar(255, 0, 0), 2);             cv::circle(imagergb, bellipse.center, 3, cv::scalar(0, 255, 0), 2);         }         //trapez         if (hull.size() == 4)         {             cv::drawcontours(imagergb, contours, i, cv::scalar(0, 255, 255), 2);             cv::circle(imagergb, bellipse.center, 3, cv::scalar(0, 0, 255), 2);         }     }     return imagergb; } 

so... in general problems coused wrong thresholding parameters, how can calculete in proper way (automatically, of course)? , how can can (lol, sorry english) prevent false positives?

thesholding - think should try otsu binarization - here theory , nice picture , here documentation. kind of thresholding trying find 2 common values in image , use average value of them threshold value.

alternatively consider using hsv color space, might easier distinguish black , white regions other regions. idea use inrange function (in rgb or in hsv color space - should work in woth situations) - need find 2 ranges (one black regions , 1 white) , search regions (using inrange function) - at post.

another way accomplish task might using library blob extraction like one or blob extractor part of opencv.

distinguish triangle trapezoid - see 2 basic ways improve solution here:

• in line cv::approxpolydp(hull, hull, 15, true); make third parameter (15 in situation) not constant value, part of contour area or length. should adapt contour size, can't canstant value. it's hard how calculate without testing - try start 1-5% of contour area or length (i start length, guess) , see whether value fine/to big/to small check other values if needed. unfortunetely there no other way, finding equation manually shouldn't take long time.
• when have 4 or 5 points calculate equations of lines join consecutive points (point 1 point 2, point 2 point 3, etc don't forget calculate line between first point , last point), check whether 2 of lines parallel (or @ least close being parallel - angle between them close 0 degress) - if find parallel lines contour trapezoid, otherwise it's triangle.