mirror
/
OpenDroneMap-ODM
kopia lustrzana https://github.com/OpenDroneMap/ODM
1
0
Forkuj 0
Kod Zgłoszenia Projekty Wydania Wiki Aktywność

Use FLANN matcher instead of brute force

pull/1493/head
usplm 2022-06-26 22:15:24 -04:00 zatwierdzone przez Piero Toffanin
rodzic d7ae810958
commit c35ab4480b
1 zmienionych plików z 19 dodań i 6 usunięć
Pokaż statystyki Ściągnij plik aktualizacji Ściągnij plik porównania Expand all files Collapse all files

25
opendm/multispectral.py
Wyświetl plik

@ -481,12 +481,12 @@ def find_features_homography(image_gray, align_image_gray, feature_retention=0.7
kp_align_image, desc_align_image = detector.detectAndCompute(align_image_gray, None) kp_align_image, desc_align_image = detector.detectAndCompute(align_image_gray, None)
# Match # Match
bf = cv2.BFMatcher(cv2.NORM_L1,crossCheck=True) # bf = cv2.BFMatcher(cv2.NORM_L1,crossCheck=True)
try: # try:
matches = bf.match(desc_image, desc_align_image) # matches = bf.match(desc_image, desc_align_image)
except Exception as e: # except Exception as e:
log.ODM_INFO("Cannot match features") # log.ODM_INFO("Cannot match features")
return None # return None
# Sort by score # Sort by score
# matches.sort(key=lambda x: x.distance, reverse=False) # matches.sort(key=lambda x: x.distance, reverse=False)
@ -495,6 +495,18 @@ def find_features_homography(image_gray, align_image_gray, feature_retention=0.7
# num_good_matches = int(len(matches) * feature_retention) # num_good_matches = int(len(matches) * feature_retention)
# matches = matches[:num_good_matches] # matches = matches[:num_good_matches]
# Use FLANN based method to match keypoints
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv2.FlannBasedMatcher(index_params, search_params)
try:
matches = flann.knnMatch(desc_image, desc_align_image, k=2)
except Exception as e:
log.ODM_INFO("Cannot match features")
return None
# Filter good matches following Lowe's ratio test # Filter good matches following Lowe's ratio test
good_matches = [] good_matches = []
for m, n in matches: for m, n in matches:
@ -510,6 +522,7 @@ def find_features_homography(image_gray, align_image_gray, feature_retention=0.7
# Debug # Debug
# imMatches = cv2.drawMatches(im1, kp_image, im2, kp_align_image, matches, None) # imMatches = cv2.drawMatches(im1, kp_image, im2, kp_align_image, matches, None)
# cv2.imwrite("matches.jpg", imMatches) # cv2.imwrite("matches.jpg", imMatches)
log.ODM_INFO("Good feature matches: %s" % len(matches))
# Extract location of good matches # Extract location of good matches
points_image = np.zeros((len(matches), 2), dtype=np.float32) points_image = np.zeros((len(matches), 2), dtype=np.float32)