def test_process_minutiae(self):
# Tests minutiae creation output.
path = '../../data/Fingerprints - Set A/101_1.tif'
img = load_image(path, True)
img = enhance_image(img, padding=5)
minutiae = process_minutiae(img)
self.assertIsNotNone(minutiae, msg='ERROR: Wrong output.')
def test_enhance_image(self):
# Tests image enhancing output and data loss.
# Revert gray colour levels. Match scale with the raw image for comparison.
path = '../../data/Fingerprints - Set A/101_1.tif'
# Image loading
img = load_image(path, True)
img = enhance_image(img, padding=5)
def trainData(self):
"""
Loads model on the given dataset.
"""
start = time.time()
print(f'INFO: Loading model features. Model: {self.model.lower()}')
if self.model.lower() == 'tree':
for i in range(len(self.images)):
# Extract minutiae.
self.images[i].image_enhanced = enhance_image(
self.images[i].image_raw, skeletonise=True)
minutiae = process_minutiae(self.images[i].image_enhanced)
# Confirmed point matching.
self.images[i].profile = generate_tuple_profile(minutiae)
# Rewriting to the loaded data.
self.images[i].minutiae = minutiae
elif self.model.lower() == 'orb':
# Base data.
print('INFO: Training skipped.')
elif self.model.lower() == 'bf':
for i in range(len(self.images)):
# BFMatcher descriptors generation.
self.images[i].image_enhanced = enhance_image(
self.images[i].image_raw, skeletonise=False)
points, descriptors = edge_processing(
self.images[i].image_enhanced, threshold=self.threshold)
# points, descriptors = minutiae_points(self.images[i].image_enhanced)
self.images[i].descriptors = descriptors
print(
f'INFO: Training completed in {round(time.time() - start, 2)} sec')
from libs.basics import load_image, display_image
from libs.minutiae import process_minutiae, generate_tuple_profile
from libs.enhancing import enhance_image
from libs.matching import match_tuples, build_edges, edge_matching
import unittest
path_base = '../../data/Fingerprints - Set A/101_2.tif'
path_test = '../../data/Fingerprints - Set A/101_2.tif'
img_base = load_image(path_base, True)
img_base = enhance_image(img_base, padding=5)
img_test = load_image(path_test, True)
img_test = enhance_image(img_test, padding=5)
# Confirmed point matching.
TUPLE_BASE = generate_tuple_profile(process_minutiae(img_base))
TUPLE_TEST = generate_tuple_profile(process_minutiae(img_test))
class TestMatching(unittest.TestCase):
def test_match_tuples(self):
# Tests minutiae creation output.
ccpb_base, ccpb_test = match_tuples(TUPLE_BASE, TUPLE_TEST)
print('---')
def test_edge_matching(self):
def matchFingerprint(self,
image: np.array,
verbose: bool = False,
match_th: int = 33):
"""
The given image is compared against the loaded templates.
A similarity score is computed and used to determine the most likely match, if any.
"""
if self.model.lower() == 'bf':
# BFMatcher and MES scorings.
scores = {}
# Returns score, aims to minimise them for computing the best match.
# Test descriptors.
img = enhance_image(image, skeletonise=False)
points, descriptors = edge_processing(img,
threshold=self.threshold)
for i in range(len(self.images)):
# Matching.
try:
matches = match_edge_descriptors(
self.images[i].descriptors, descriptors)
except AttributeError:
raise Exception(
'ERROR: Model not trained - run trainData first.')
# Calculate score
score = sum([match.distance for match in matches])
# Using mes (mean edge score) = sum(match distance) / len(matches)
if len(matches) > 0:
mes = score / len(matches)
scores[self.images[i].img_id] = mes
scores = sorted(scores.items(), key=operator.itemgetter(1))
# Display most likely match
results = [{
'img_id': s[0],
'score': round(s[1], 2),
'match': s[1] < match_th
} for s in scores]
matches = [m for m in results if m['match']]
if len(matches) == 0:
print(
f'No match found. Most similar fingerprint is {results[:5]}'
)
else:
print(f'INFO: Matches found, score: {matches}')
return scores
elif self.model.lower() == 'orb':
# Basic SIFT based ORB matcher.
for i in self.images:
sift_match(i.image_raw, image)
elif self.model.lower() == 'tree':
img_test = enhance_image(image, skeletonise=True)
minutiae_test = process_minutiae(img_test)
# Confirmed point matching.
img_profile = generate_tuple_profile(minutiae_test)
for i in range(len(self.images)):
# Matching.
common_points_base, common_points_test = match_tuples(
self.images[i].profile, img_profile)
if evaluate(common_points_base, self.images[i].minutiae,
minutiae_test):
print(f'Match with {self.images[i].img_id}')
else:
print(f'Not a match with {self.images[i].img_id}')
elif self.model == 'cnn':
pass
else:
print('INFO: Not implemented yet.')