def _tefi(self, note, second_grade_candidates):
if self.debug:
print 'Performing tefi'
rotated_templates = []
for cshift in range(self.settings['nr_of_rotation_angles']):
rotation_matrix = cv2.getRotationMatrix2D((self.notesize / 2, self.notesize / 2), -cshift * 10, 1)
rotated_note = cv2.warpAffine(note, rotation_matrix, (self.notesize, self.notesize))
masked_rotated_note = flatten_disc(rotated_note, (self.notesize / 2, self.notesize / 2), self.notesize / 2)
rotated_templates.append(masked_rotated_note)
candidates_with_correlation = []
for x, y, cshift in second_grade_candidates:
for x2 in [x - 1, x, x + 1]:
for y2 in [y - 1, y, y + 1]:
possible_match = imagefuncs.submatrix(self.board, x2, y2, self.notesize)
masked_possible_match = flatten_disc(possible_match, (self.notesize / 2, self.notesize / 2), self.notesize / 2)
corr = calculate_correlation(rotated_templates[cshift],
masked_possible_match,
self.settings['thresh_contrast'],
self.settings['thresh_brightness'])
#print 'correlation with second grade candidate at (%d,%d) is %f' % (x2, y2, corr)
candidates_with_correlation.append( (corr, x2, y2) )
candidates_with_correlation.sort(reverse = True)
final_match = candidates_with_correlation[0]
if final_match[0] < self.settings['thresh_confidence']:
if self.debug:
print 'no match was found. The best was at (%d,%d) with correlation %f' % (final_match[1], final_match[2], final_match[0])
imagefuncs.qimshow(imagefuncs.submatrix(self.board, final_match[1], final_match[2], self.notesize))
return None
else:
if self.debug:
print 'final match is at (%d,%d) with correlation %f' % (final_match[1], final_match[2], final_match[0])
imagefuncs.qimshow(imagefuncs.submatrix(self.board, final_match[1], final_match[2], self.notesize))
for corr, x, y in candidates_with_correlation:
diff_x = x - final_match[1]
diff_y = y - final_match[2]
if diff_x * diff_x + diff_y * diff_y > self.notesize * self.notesize:
print 'second best match is at (%d,%d) with correlation %f' % (x, y, corr)
imagefuncs.qimshow(imagefuncs.submatrix(self.board, x, y, self.notesize))
break
return final_match
def findsquares(scrumboard, image):
start_of_done, _ = scrumboard.get_position_from_state('done')
masked_image = image[:,:start_of_done,:]
#imagefuncs.qimshow(masked_image)
notepositions = findnotes(masked_image)
squares = []
for pos in notepositions:
notebitmap = imagefuncs.submatrix(image, pos[0], pos[1], imagefuncs.NOTE_SIZE)
squares.append(board.Square(notebitmap, (pos[0], pos[1])))
#imagefuncs.qimshow(['found square', notebitmap])
return squares
def _rafi(self, note, first_grade_candidates):
if self.debug:
print 'Performing rafi'
rafi_masks = []
rafi_means = []
for i in range(self.settings['nr_of_rotation_angles']):
angle = -(math.pi / 2) + (i * 2 * math.pi) / 36
halfnotesize = self.notesize / 2
p1 = (halfnotesize, halfnotesize)
p2 = (int(halfnotesize + halfnotesize * math.cos(angle)),
int(halfnotesize + halfnotesize * math.sin(angle)))
mask = np.zeros((self.notesize, self.notesize), dtype=np.uint8)
cv2.line(mask, p1, p2, (255,255,255))
rafi_masks.append(mask)
rafi_means.append(cv2.mean(note, mask)[0])
rafi_means = np.array(rafi_means)
candidate_means = []
for x, y in first_grade_candidates:
candidate = imagefuncs.submatrix(self.board, x, y, self.notesize)
this_candidate_means = []
for mask in rafi_masks:
this_candidate_means.append(cv2.mean(candidate, mask)[0])
candidate_means.append(this_candidate_means)
candidates_with_correlation = []
for i, this_candidate_means in enumerate(candidate_means):
x, y = first_grade_candidates[i]
this_candidate_rascorrs = []
max_rascorr = -2
max_cshift = None
for cshift in range(self.settings['nr_of_rotation_angles']):
rotated_rafi_means = np.roll(rafi_means, cshift)
rascorr = calculate_correlation(np.array(this_candidate_means),
rotated_rafi_means,
self.settings['thresh_contrast'],
self.settings['thresh_brightness'])
this_candidate_rascorrs.append(rascorr)
if rascorr > max_rascorr:
max_rascorr = rascorr
max_cshift = cshift
candidates_with_correlation.append( (max_rascorr, x, y, max_cshift) )
nr_of_second_grade_candidates = int((len(candidates_with_correlation) * self.settings['percentage_of_second_grade_candidates']) / 100)
if nr_of_second_grade_candidates < self.settings['min_nr_of_second_grade_candidates']:
nr_of_second_grade_candidates = min(len(candidates_with_correlation), self.settings['min_nr_of_second_grade_candidates'])
second_grade_candidates = sorted([ (x, y, cshift) for _, x, y, cshift in
sorted(candidates_with_correlation, reverse=True)[:nr_of_second_grade_candidates] ])
return second_grade_candidates
