masking out the field somewhat more intelligently

pykick/detection_demo.py

@@ -104,8 +104,7 @@ if __name__ == '__main__':
                 cv2.bitwise_and(frame, frame, mask=field))
             goal_finder.draw(frame, goal)
             ball_finder.draw(frame, ball)
-            cv2.imshow(window_name,
-                cv2.bitwise_and(frame, frame, mask=field))
+            cv2.imshow(window_name, frame)
 
             key = cv2.waitKey(0 if args.manual else 1)
             if key == ord('q') or key == 27:

pykick/utils.py

@@ -4,6 +4,7 @@ import os
 import json
 
 import cv2
+import numpy as np
 
 
 HERE = os.path.dirname(os.path.realpath(__file__))
@@ -34,8 +35,17 @@ def field_mask(frame, hsv_lower, hsv_upper):
     hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
     blurred = cv2.GaussianBlur(hsv, (25, 25), 20)
     thr = cv2.inRange(blurred, tuple(hsv_lower), tuple(hsv_upper))
+    thr = cv2.erode(thr, None, iterations=4)
+    thr = cv2.dilate(thr, None, iterations=8)
+    cnts, _ = cv2.findContours(thr.copy(), cv2.RETR_EXTERNAL,
+                               cv2.CHAIN_APPROX_SIMPLE)
+    field = max(cnts, key=cv2.contourArea)
+    field = cv2.convexHull(field)
+    # print(field)
+    mask = np.zeros(thr.shape, dtype=np.uint8)
+    print(mask.dtype)
+    thr = cv2.cvtColor(thr, cv2.COLOR_GRAY2BGR)
+    cv2.drawContours(mask, (field,), -1, 255, -1)
 
     # The ususal
-    thr = cv2.erode(thr, None, iterations=6)
-    thr = cv2.dilate(thr, None, iterations=20)
-    return thr
+    return mask