def identify(self, bounding_box_content: numpy.ndarray, mask: numpy.ndarray=None) -> Subject:
bounding_box_content, mask = self.__extract_person(bounding_box_content, mask)
image_hsv = cv2.cvtColor(bounding_box_content, cv2.COLOR_BGR2HSV)
hue_channel, saturation_channel, _ = cv2.split(image_hsv)
hue_images = self.__split_image(hue_channel)
sat_images = self.__split_image(saturation_channel)
image_bgr = bounding_box_content
blue_channel, green_channel, red_channel = cv2.split(image_bgr)
blue_images = self.__split_image(blue_channel)
green_images = self.__split_image(green_channel)
red_images = self.__split_image(red_channel)
image_ycrcb = cv2.cvtColor(bounding_box_content, cv2.COLOR_BGR2YCrCb)
_, cr_channel, cb_channel = cv2.split(image_ycrcb)
cr_images = self.__split_image(cr_channel)
cb_images = self.__split_image(cb_channel)
masks = self.__split_image(mask)
hue_histograms = self.__calculate_histograms(masks, 19, (0, 180), hue_images)
sat_histograms = self.__calculate_histograms(masks, 16, (0, 256), sat_images)
blue_histograms = self.__calculate_histograms(masks, 16, (0, 256), blue_images)
green_histograms = self.__calculate_histograms(masks, 16, (0, 256), green_images)
red_histograms = self.__calculate_histograms(masks, 16, (0, 256), red_images)
cr_histograms = self.__calculate_histograms(masks, 16, (0, 256), cr_images)
cb_histograms = self.__calculate_histograms(masks, 16, (0, 256), cb_images)
new_histograms = SubjectHistograms(hue_histograms, sat_histograms, blue_histograms, green_histograms,
red_histograms, cr_histograms, cb_histograms)
self.__new_histograms = new_histograms
if self.__feature_matching_enabled:
new_descriptors = self.__get_descriptors((image_bgr, mask))
else:
new_descriptors = None
if self.__add_to_db:
if self.__subject is None:
self.__subject = Subject(self.__add_to_db, new_histograms, new_descriptors)
Database.add_subject(self.__subject)
else:
self.__subject.add_histograms(new_histograms)
if self.__feature_matching_enabled:
self.__subject.add_descriptors(new_descriptors)
for subject in self.__subjects:
overall_score = subject.check_similarities(new_histograms)
if self.__feature_matching_enabled:
self.__do_feature_match(subject.descriptors[0], new_descriptors)
try:
self.__scores[subject.name][1] += overall_score
except KeyError:
self.__scores[subject.name] = {}
self.__scores[subject.name][0], self.__scores[subject.name][1] = subject, overall_score
if self.__best_match is None or self.__scores[subject.name][1] > self.__scores[self.__best_match.name][1]:
self.__best_match = subject
if self.__feature_matching_enabled:
ordered_scores = OrderedDict(sorted(self.__scores.items(), key=lambda t: t[1][1], reverse=True))
counter = 0
highest_feature_match = 0
for subject_with_score in ordered_scores.values():
feature_match = self.__do_feature_match(subject_with_score[0].descriptors[0], new_descriptors)
print(subject_with_score[0].name, feature_match)
if feature_match > highest_feature_match:
self.__best_match = subject_with_score[0]
highest_feature_match = feature_match
if counter > 8:
break
counter += 1
return self.__best_match
