find_obj.py allows to select detector and matcher and explore matching pairs

samples/python2/find_obj.py

@@ -2,18 +2,62 @@
 Feature-based image matching sample.
 
 USAGE
-     find_obj.py [ <image1> <image2> ]
+  find_obj.py [--feature=<sift|surf|orb>[-flann]] [ <image1> <image2> ]
+
+  --feature  - Feature to use. Can be sift, surf of orb. Append '-flann' to feature name
+                to use Flann-based matcher instead bruteforce.
+
+  Press left mouse button on a feature point to see its mathcing point.
+
 '''
 
 import numpy as np
 import cv2
+from common import anorm, getsize
 
 FLANN_INDEX_KDTREE = 1  # bug: flann enums are missing
-
+FLANN_INDEX_LSH    = 6
-flann_params = dict(algorithm = FLANN_INDEX_KDTREE, trees = 5)
 
 
-def draw_match(img1, img2, p1, p2, status = None, H = None):
+def init_feature(name):
+    detector, matcher = None, None
+    chunks = name.split('-')
+    if chunks[0] == 'sift':
+        detector = cv2.SIFT()
+        norm = cv2.NORM_L2
+    elif chunks[0] == 'surf':
+        detector = cv2.SURF(1000)
+        norm = cv2.NORM_L2
+    elif chunks[0] == 'orb':
+        detector = cv2.ORB(500)
+        norm = cv2.NORM_HAMMING
+    if 'flann' in chunks:
+        if norm == cv2.NORM_L2:
+            flann_params = dict(algorithm = FLANN_INDEX_KDTREE, trees = 5)
+        else:
+            flann_params= dict(algorithm = FLANN_INDEX_LSH,
+                               table_number = 6, # 12
+                               key_size = 12,     # 20
+                               multi_probe_level = 1) #2
+        matcher = cv2.FlannBasedMatcher(flann_params, {})  # bug : need to pass empty dict (#1329)
+    else:    
+        matcher = cv2.BFMatcher(norm)
+    return detector, matcher
+
+    
+def filter_matches(kp1, kp2, matches, ratio = 0.75):
+    mkp1, mkp2 = [], []
+    for m in matches:
+        if len(m) == 2 and m[0].distance < m[1].distance * ratio:
+            m = m[0]
+            mkp1.append( kp1[m.queryIdx] )
+            mkp2.append( kp2[m.trainIdx] )
+    p1 = np.float32([kp.pt for kp in mkp1])
+    p2 = np.float32([kp.pt for kp in mkp2])
+    kp_pairs = zip(mkp1, mkp2)
+    return p1, p2, kp_pairs
+   
+def explore_match(win, img1, img2, kp_pairs, status = None, H = None):
     h1, w1 = img1.shape[:2]
     h2, w2 = img2.shape[:2]
     vis = np.zeros((max(h1, h2), w1+w2), np.uint8)
@@ -25,71 +69,96 @@ def draw_match(img1, img2, p1, p2, status = None, H = None):
         corners = np.float32([[0, 0], [w1, 0], [w1, h1], [0, h1]])
         corners = np.int32( cv2.perspectiveTransform(corners.reshape(1, -1, 2), H).reshape(-1, 2) + (w1, 0) )
         cv2.polylines(vis, [corners], True, (255, 255, 255))
-    
+
     if status is None:
         status = np.ones(len(p1), np.bool_)
+    p1 = np.int32([kpp[0].pt for kpp in kp_pairs])
+    p2 = np.int32([kpp[1].pt for kpp in kp_pairs]) + (w1, 0)
+
     green = (0, 255, 0)
     red = (0, 0, 255)
-    for (x1, y1), (x2, y2), inlier in zip(np.int32(p1), np.int32(p2), status):
+    white = (255, 255, 255)
-        col = [red, green][inlier]
+    kp_color = (51, 103, 236)
+    for (x1, y1), (x2, y2), inlier in zip(p1, p2, status):
         if inlier:
-            cv2.line(vis, (x1, y1), (x2+w1, y2), col)
+            col = green
             cv2.circle(vis, (x1, y1), 2, col, -1)
-            cv2.circle(vis, (x2+w1, y2), 2, col, -1)
+            cv2.circle(vis, (x2, y2), 2, col, -1)
         else:
+            col = red
             r = 2
             thickness = 3
             cv2.line(vis, (x1-r, y1-r), (x1+r, y1+r), col, thickness)
             cv2.line(vis, (x1-r, y1+r), (x1+r, y1-r), col, thickness)
-            cv2.line(vis, (x2+w1-r, y2-r), (x2+w1+r, y2+r), col, thickness)
+            cv2.line(vis, (x2-r, y2-r), (x2+r, y2+r), col, thickness)
-            cv2.line(vis, (x2+w1-r, y2+r), (x2+w1+r, y2-r), col, thickness)
+            cv2.line(vis, (x2-r, y2+r), (x2+r, y2-r), col, thickness)
-    return vis
+    vis0 = vis.copy()
+    for (x1, y1), (x2, y2), inlier in zip(p1, p2, status):
+        if inlier:
+            cv2.line(vis, (x1, y1), (x2, y2), green)
+
+    cv2.imshow(win, vis)
+    def onmouse(event, x, y, flags, param):
+        cur_vis = vis
+        if flags & cv2.EVENT_FLAG_LBUTTON:
+            cur_vis = vis0.copy()
+            r = 8
+            m = (anorm(p1 - (x, y)) < r) | (anorm(p2 - (x, y)) < r)
+            idxs = np.where(m)[0]
+            kp1s, kp2s = [], []
+            for i in idxs:
+                 (x1, y1), (x2, y2) = p1[i], p2[i]
+                 col = (red, green)[status[i]]
+                 cv2.line(cur_vis, (x1, y1), (x2, y2), col)
+                 kp1, kp2 = kp_pairs[i]
+                 kp1s.append(kp1)
+                 kp2s.append(kp2)
+            cur_vis = cv2.drawKeypoints(cur_vis, kp1s, flags=4, color=kp_color)
+            cur_vis[:,w1:] = cv2.drawKeypoints(cur_vis[:,w1:], kp2s, flags=4, color=kp_color)
+
+        cv2.imshow(win, cur_vis)
+    cv2.setMouseCallback(win, onmouse)
 
 
 if __name__ == '__main__':
     print __doc__
     
-    import sys
+    import sys, getopt
-    try: fn1, fn2 = sys.argv[1:3]
+    opts, args = getopt.getopt(sys.argv[1:], '', ['feature='])
+    opts = dict(opts)
+    feature_name = opts.get('--feature', 'sift')
+    try: fn1, fn2 = args
     except:
         fn1 = '../c/box.png'
         fn2 = '../c/box_in_scene.png'
-
+        
     img1 = cv2.imread(fn1, 0)
     img2 = cv2.imread(fn2, 0)
+    detector, matcher = init_feature(feature_name)
+    if detector != None:
+        print 'using', feature_name
+    else:
+        print 'unknown feature:', feature_name
+        sys.exit(1)
+
 
-    detector = cv2.SIFT()
     kp1, desc1 = detector.detectAndCompute(img1, None)
     kp2, desc2 = detector.detectAndCompute(img2, None)
     print 'img1 - %d features, img2 - %d features' % (len(kp1), len(kp2))
 
-    bf_matcher = cv2.BFMatcher(cv2.NORM_L2)
+    def match_and_draw(win):
-    flann_matcher = cv2.FlannBasedMatcher(flann_params, {})  # bug : need to pass empty dict (#1329)
+        print 'matching...'
-
+        raw_matches = matcher.knnMatch(desc1, trainDescriptors = desc2, k = 2) #2
-    def match_and_draw(matcher, r_threshold = 0.75):
+        p1, p2, kp_pairs = filter_matches(kp1, kp2, raw_matches)
-        raw_matches = matcher.knnMatch(desc1, trainDescriptors = desc2, k = 2)
-        p1, p2 = [], []
-        for m in raw_matches:
-            if len(m) == 2 and m[0].distance < m[1].distance * r_threshold:
-                m = m[0]
-                p1.append( kp1[m.queryIdx].pt )
-                p2.append( kp2[m.trainIdx].pt )
-        p1, p2 = np.float32((p1, p2))
         if len(p1) >= 4:
-            H, status = cv2.findHomography(p1, p2, cv2.RANSAC, 2.0)
+            H, status = cv2.findHomography(p1, p2, cv2.RANSAC, 5.0)
             print '%d / %d  inliers/matched' % (np.sum(status), len(status))
         else:
             H, status = None, None
             print '%d matches found, not enough for homography estimation' % len(p1)
 
-        vis = draw_match(img1, img2, p1, p2, status, H)
+        vis = explore_match(win, img1, img2, kp_pairs, status, H)
-        return vis
 
-    print 'bruteforce match:',
+    match_and_draw('find_obj')
-    vis_brute = match_and_draw( bf_matcher )
+    cv2.waitKey()
-    print 'flann match:',
-    vis_flann = match_and_draw( flann_matcher )
-    cv2.imshow('find_obj', vis_brute)
-    cv2.imshow('find_obj flann', vis_flann)
-    0xFF & cv2.waitKey()
     cv2.destroyAllWindows()