Added support for colored circles recognition instead of threshold-by…

…-color previous method

cdect/README.mdwn

@@ -0,0 +1,2 @@
+This component is dedicated to find cirlces using Python + opencv
+

cdect/c.txt

@@ -0,0 +1,2 @@
+543	188
+516	645

cdect/cdect.py

@@ -0,0 +1,85 @@
+import cv2
+import cv
+import numpy as np
+
+def get_circles(img, args):
+    # TODO: for param2 in range(args.max_param2, args.min_param2, -1)
+    circles = cv2.HoughCircles(img,cv.CV_HOUGH_GRADIENT,1,
+                               args.min_dist,
+                               param2=args.param2,
+                               minRadius=args.min_radius,maxRadius=args.max_radius)
+    if circles is None:
+        raise Exception('error looking for circles')
+
+    return circles
+
+
+def highlight(img, circles, args):
+    for i in circles[0,:]:
+        # draw the outer circle
+        cv2.circle(img,(i[0],i[1]),i[2],(255,255,0),2)
+        # draw the center of the circle
+        cv2.circle(img,(i[0],i[1]),2,(255,0,255),3)
+    return img
+
+def main(args):
+    img = cv2.imread(args.image)
+    cimg = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+    if False:
+        cv2.imshow('hsv', cv2.cvtColor(cimg, cv2.COLOR_HSV2BGR))
+        cv2.waitKey(0)
+        cv2.destroyAllWindows()
+
+    lower = np.array([args.hue - args.hue_radius, args.min_saturation, args.min_value])
+    upper = np.array([args.hue + args.hue_radius,255,args.max_value])
+    lower = np.array(lower, dtype = "uint8")
+    upper = np.array(upper, dtype = "uint8")
+    # find the colors within the specified boundaries and apply
+    # the mask
+    mask = cv2.inRange(cimg, lower, upper)
+    cimg = cv2.bitwise_and(cimg, cimg, mask = mask)
+    if True:
+        cv2.imshow('threshold', cv2.cvtColor(cimg, cv2.COLOR_HSV2BGR))
+        cv2.waitKey(0)
+        cv2.destroyAllWindows()
+    gray = cv2.cvtColor(cv2.cvtColor(cimg, cv2.COLOR_HSV2BGR),
+                        cv2.COLOR_BGR2GRAY)
+    cv2.destroyAllWindows()
+    circles = get_circles(gray, args)
+    if args.coordinates:
+        with open(args.coordinates, 'w') as buf:
+            for c in circles[0]:
+                p1, p2, r = c
+                buf.write('%d\t%d\n' % (p1, p2))
+
+    if args.show_result:
+        highlight(img, circles, args)
+        cv2.imshow('detected circles',img)
+        cv2.waitKey(0)
+        cv2.destroyAllWindows()
+
+def get_parser():
+    import argparse
+    p = argparse.ArgumentParser()
+    p.add_argument('image')
+
+    col = p.add_argument_group('color')
+    col.add_argument('--hue', default=59, type=int)
+    col.add_argument('--hue-radius', default=10, type=int)
+    col.add_argument('--min-saturation', default=150, type=int)
+    col.add_argument('--min-value', default=150, type=int)
+    col.add_argument('--max-value', default=255, type=int)
+
+    det = p.add_argument_group('circle detection')
+    det.add_argument('--param2', default=20, type=int, help='If it is lower, it will have more false positives')
+    det.add_argument('--min-dist', default=100, type=int, help='Minimum distance between circles')
+    det.add_argument('--min-radius', default=0, type=int)
+    det.add_argument('--max-radius', default=0, type=int)
+
+    p.add_argument('--coordinates', metavar='FILE', help='Write found coordinates to FILE')
+    p.add_argument('--show-result', action='store_true', default=False, help='Show image with found circles')
+
+    return p
+
+if __name__ == '__main__':
+    main(get_parser().parse_args())