- labeling algorithm
- largest from labeling
- rotate the image. labeling does not work.. hmmm..
next try:
-
color segmenting
-
find contours from masked
-
find corners.. how? using opencv feature detection or else.. or simply sorting pixels based on coordinates..
-
find ends.. hmm how. two closest pairs from the corner coordinates because the table is not a square
-
remove find contours from masked, otherwise same
That did not work if the table was too straight because upper most pixel
-
Find
rect = cv2.minAreaRect(non_zero_pixels) points = cv2.cv.BoxPoints(rect) points = np.int0(np.around(points))
and remove own hacky find corners function.
-
This leads to pretty solid table end detection and is fast because internal functions
-
next challenge: how to get position of score dots? try: half of corner dots, and half again