|
|
|
import numpy as np
|
|
|
|
import cv2
|
|
|
|
from common import anorm
|
|
|
|
from functools import partial
|
|
|
|
|
|
|
|
help_message = '''SURF image match
|
|
|
|
|
|
|
|
USAGE: findobj.py [ <image1> <image2> ]
|
|
|
|
'''
|
|
|
|
|
|
|
|
FLANN_INDEX_KDTREE = 1 # bug: flann enums are missing
|
|
|
|
|
|
|
|
flann_params = dict(algorithm = FLANN_INDEX_KDTREE,
|
|
|
|
trees = 4)
|
|
|
|
|
|
|
|
def match_bruteforce(desc1, desc2, r_threshold = 0.75):
|
|
|
|
res = []
|
|
|
|
for i in xrange(len(desc1)):
|
|
|
|
dist = anorm( desc2 - desc1[i] )
|
|
|
|
n1, n2 = dist.argsort()[:2]
|
|
|
|
r = dist[n1] / dist[n2]
|
|
|
|
if r < r_threshold:
|
|
|
|
res.append((i, n1))
|
|
|
|
return np.array(res)
|
|
|
|
|
|
|
|
def match_flann(desc1, desc2, r_threshold = 0.6):
|
|
|
|
flann = cv2.flann_Index(desc2, flann_params)
|
|
|
|
idx2, dist = flann.knnSearch(desc1, 2, params = {}) # bug: need to provide empty dict
|
|
|
|
mask = dist[:,0] / dist[:,1] < r_threshold
|
|
|
|
idx1 = np.arange(len(desc1))
|
|
|
|
pairs = np.int32( zip(idx1, idx2[:,0]) )
|
|
|
|
return pairs[mask]
|
|
|
|
|
|
|
|
def draw_match(img1, img2, p1, p2, status = None, H = None):
|
|
|
|
h1, w1 = img1.shape[:2]
|
|
|
|
h2, w2 = img2.shape[:2]
|
|
|
|
vis = np.zeros((max(h1, h2), w1+w2), np.uint8)
|
|
|
|
vis[:h1, :w1] = img1
|
|
|
|
vis[:h2, w1:w1+w2] = img2
|
|
|
|
vis = cv2.cvtColor(vis, cv2.COLOR_GRAY2BGR)
|
|
|
|
|
|
|
|
if H is not 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_)
|
|
|
|
green = (0, 255, 0)
|
|
|
|
red = (0, 0, 255)
|
|
|
|
for (x1, y1), (x2, y2), inlier in zip(np.int32(p1), np.int32(p2), status):
|
|
|
|
col = [red, green][inlier]
|
|
|
|
if inlier:
|
|
|
|
cv2.line(vis, (x1, y1), (x2+w1, y2), col)
|
|
|
|
cv2.circle(vis, (x1, y1), 2, col, -1)
|
|
|
|
cv2.circle(vis, (x2+w1, y2), 2, col, -1)
|
|
|
|
else:
|
|
|
|
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+w1-r, y2+r), (x2+w1+r, y2-r), col, thickness)
|
|
|
|
return vis
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
import sys
|
|
|
|
try: fn1, fn2 = sys.argv[1:3]
|
|
|
|
except:
|
|
|
|
fn1 = '../c/box.png'
|
|
|
|
fn2 = '../c/box_in_scene.png'
|
|
|
|
print help_message
|
|
|
|
|
|
|
|
img1 = cv2.imread(fn1, 0)
|
|
|
|
img2 = cv2.imread(fn2, 0)
|
|
|
|
|
|
|
|
surf = cv2.SURF(1000)
|
|
|
|
kp1, desc1 = surf.detectAndCompute(img1, None)
|
|
|
|
kp2, desc2 = surf.detectAndCompute(img2, None)
|
|
|
|
desc1.shape = (-1, surf.descriptorSize())
|
|
|
|
desc2.shape = (-1, surf.descriptorSize())
|
|
|
|
print 'img1 - %d features, img2 - %d features' % (len(kp1), len(kp2))
|
|
|
|
|
|
|
|
def match_and_draw(match, r_threshold):
|
|
|
|
m = match(desc1, desc2, r_threshold)
|
|
|
|
matched_p1 = np.array([kp1[i].pt for i, j in m])
|
|
|
|
matched_p2 = np.array([kp2[j].pt for i, j in m])
|
|
|
|
H, status = cv2.findHomography(matched_p1, matched_p2, cv2.RANSAC, 5.0)
|
|
|
|
print '%d / %d inliers/matched' % (np.sum(status), len(status))
|
|
|
|
|
|
|
|
vis = draw_match(img1, img2, matched_p1, matched_p2, status, H)
|
|
|
|
return vis
|
|
|
|
|
|
|
|
print 'bruteforce match:',
|
|
|
|
vis_brute = match_and_draw( match_bruteforce, 0.75 )
|
|
|
|
print 'flann match:',
|
|
|
|
vis_flann = match_and_draw( match_flann, 0.6 ) # flann tends to find more distant second
|
|
|
|
# neighbours, so r_threshold is decreased
|
|
|
|
cv2.imshow('find_obj SURF', vis_brute)
|
|
|
|
cv2.imshow('find_obj SURF flann', vis_flann)
|
|
|
|
0xFF & cv2.waitKey()
|
|
|
|
cv2.destroyAllWindows()
|