-
Notifications
You must be signed in to change notification settings - Fork 0
/
feat_matching.py
279 lines (246 loc) · 9.07 KB
/
feat_matching.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Use ORB to set keypoints, then features matching, then Homograhy,
# then matchlines
import sys
import numpy as np
import cv2
ver = (cv2.__version__)
print( "Open CV Version",ver )
GOOD_SUM_THRES = 100
GOOD_THRES = 3
orb_win = "ORB Matching"
cv2.namedWindow( orb_win )
cube_names = ["tromb", "lamp", "baby"]
cube_colors = [(255,0,0), (0,255,0), (0,0,255)]
cube_size = 124
## info on cubes
cube = {}
cube_gray = {}
cube_kp = {}
cube_desc={}
## info on unknown image
ukn_img = None
k, desc = None, None
# read images
for n in cube_names:
cube[n] = cv2.imread( "Images/cu_{}_{}.png".format( n, cube_size ) )
if( cube[n] is None):
print( "Could not read Images/cu_{}_{}.png".format( n, cube_size ))
sys.exit()
cube_gray[n] = np.uint8(cv2.cvtColor( cube[n], cv2.COLOR_RGB2GRAY))
ukn_img = cv2.imread( "Images/cu_baby_320.jpg" )
if( ukn_img is None):
print( "Could not read Images/cu_baby_320.jpg" )
sys.exit()
ukn_gray = np.uint8(cv2.cvtColor( ukn_img, cv2.COLOR_RGB2GRAY))
# basic default image
# new image
height = max(3*cube_size, ukn_img.shape[0] )
width = 2*cube_size + ukn_img.shape[1]
base_img = np.zeros( (height, width, 3), np.uint8 )
base_img[:, 0:cube_size] = 255
# _idc = 0
# for n in cube_names:
# base_img[_idc*cube_size:(_idc+1)*cube_size,0:cube_size] = cube[n]
# _idc += 1
#ORB detector
orb = cv2.ORB_create()
#cur_patch_size = orb.getPatchSize()
cur_patch_size = 10
orb.setPatchSize( cur_patch_size )
orb.setEdgeThreshold( cur_patch_size )
max_patch_size = 100
#cur_fast_thres = orb.getFastThreshold()
cur_fast_thres = 90
orb.setFastThreshold( cur_fast_thres )
max_fast_thres = 100
cur_scale_factor = orb.getScaleFactor()
max_scale_factor = 10 # between 1 and 2
cur_nlevels = orb.getNLevels()
max_nlevels = 10
cur_first_lvl = orb.getFirstLevel()
max_first_level = 10
# display globals
nb_lines = 10
max_nb_lines = 100
id_cube = 4
max_id_cube = len(cube_names)+1
cv_font = cv2.FONT_HERSHEY_SIMPLEX
fontscale = 0.5
def orb_and_match():
# cubes
cube_kp = {}
cube_desc={}
for n in cube_names:
cube_kp[n] = orb.detect( cube_gray[n], None )
cube_kp[n], cube_desc[n] = orb.compute( cube_gray[n], cube_kp[n] )
print( "Found {} kp for cube {}".format( len(cube_kp[n]), n))
# unknown
kp = orb.detect( ukn_gray, None )
kp, desc = orb.compute( ukn_gray, kp )
print( "Found {} kp for cube {}".format( len(kp), "Images/cu_baby_320.jpg"))
# Look for best Match Brute Force and FLANN
bforce = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
# FLANN_INDEX_KDTREE = 8
# FLANN_INDEX_LSH = 6
# index_params = dict( algorithm = FLANN_INDEX_LSH,
# table_number = 6, #12,6
# key_size = 12, #20,12
# multi_probe_level = 1 #2,1
# )
# #search_params = dict(checks = 50)
# search_params = {}
# flann = cv2.FlannBasedMatcher(index_params, search_params)
matches_bf = {}
matches_flann = {}
good_flann = {}
best_match = None
nb_match = 0
for n in cube_names:
print( "__MATCHES for {}".format(n))
matches_bf[n] = bforce.match( cube_desc[n], desc)
matches_bf[n] = sorted(matches_bf[n], key = lambda x:x.distance)
print( "BF: Found {} matches".format( len(matches_bf[n] )))
for m in matches_bf[n]:
#print( " {} of type={}".format( m, type(m)))
print( " d={} : {} -> {} ".format( m.distance, m.queryIdx, m.trainIdx))
# matches_flann[n] = flann.knnMatch( cube_desc[n], desc, k=2)
# print( "FLANN: Found {} matches".format( len(matches_flann[n] )))
# good_flann[n] = []
# for m1,m2 in matches_flann[n]:
# print( " d={} : {} -> {} VS d={} : {} -> {} ".format(
# m1.distance, m1.queryIdx, m1.trainIdx,
# m2.distance, m2.queryIdx, m2.trainIdx
# ))
# if m1.distance < 0.7 * m2.distance:
# good_flann[n].append( m1 )
# print( "FLANN: Found {} good for {}".format( len(good_flann[n]), n))
# new image
res_img = base_img.copy()
idc = 0
for n in cube_names:
orb_img = cv2.drawKeypoints( cube[n], cube_kp[n], np.array([]),
color=cube_colors[idc], flags=0)
res_img[idc*cube_size:(idc+1)*cube_size,cube_size:2*cube_size] = orb_img
idc += 1
orb_img = cv2.drawKeypoints(ukn_img, kp, np.array([]), color=(0,255,0), flags=0)
res_img[0:ukn_img.shape[0],2*cube_size:] = orb_img
# draw lines to matches
ukn_offset = np.array( (2*cube_size, 0) )
nb_good = {}
sum_good = {}
idc = 0
for n in cube_names:
cube_offset = np.array( (cube_size, idc*cube_size) )
count_line = 0
min_dist = sys.float_info.max
sum_dist = 0
nb_good[n] = 0
sum_good[n] = 0
for m in matches_bf[n]:
if count_line < nb_lines:
pt_cube = cube_kp[n][m.queryIdx].pt
pt_ukn = kp[m.trainIdx].pt
#print( "cube {} => ukn {}".format( pt_cube, pt_ukn ))
pt_cube = np.array( pt_cube ) + cube_offset
pt_ukn = np.array( pt_ukn) + ukn_offset
#print( "__line {} -> {}".format( pt_cube, pt_ukn))
if id_cube == 4 or idc == (id_cube-1):
cv2.line( res_img,
tuple(pt_cube.astype(int)),
tuple(pt_ukn.astype(int)),
cube_colors[idc], 1 )
sum_dist += m.distance
if sum_dist < GOOD_SUM_THRES:
nb_good[n] += 1
sum_good[n] = sum_dist
if m.distance < min_dist:
min_dist = m.distance
count_line += 1
min_txt = "m: {:6.2f}".format( min_dist )
sum_txt = "s: {:6.2f}".format( sum_dist )
tw, th = cv2.getTextSize( min_txt, cv_font, fontscale, 1)[0]
cv2.putText( res_img, min_txt,
(10, 10+idc*cube_size+th),
cv_font, fontscale, cube_colors[idc], thickness = 1 )
cv2.putText( res_img, sum_txt,
(10, 10+idc*cube_size+2*th+5),
cv_font, fontscale, cube_colors[idc], thickness = 1 )
idc += 1
best_cube = None
best_nb_good = GOOD_THRES
best_sum = sys.float_info.max
for n in cube_names:
if nb_good[n] >= best_nb_good and sum_good[n] < best_sum:
best_cube = n
best_nb_good = nb_good[n]
best_sum = sum_good[n]
best_txt = "None"
if best_cube is not None:
best_txt = "*{}*".format( best_cube )
tw, th = cv2.getTextSize( best_txt, cv_font, fontscale, 2)[0]
w = res_img.shape[1]
print( "w={}, tw={}".format( w, tw ))
cv2.putText( res_img, best_txt,
(int(w - tw - 10), 10+th),
cv_font, fontscale, (0,255,0), thickness = 2 )
cv2.imshow( orb_win, res_img )
def update_patch_size( val ):
global cur_patch_size
cur_patch_size = val
orb.setPatchSize( val )
orb.setEdgeThreshold( val )
orb_and_match()
def update_fast_treshold( val ):
global cur_fast_thres
cur_fast_thres = val
orb.setFastThreshold( val )
orb_and_match()
def update_scale_factor( val ):
cur_scale_factor = 1.0 + val / 10.0
orb.setScaleFactor( cur_scale_factor )
orb_and_match()
def update_nlevels( val ):
global cur_nlevels, cur_first_lvl
cur_nlevels = val
if cur_first_lvl >= cur_nlevels:
cur_first_lvl = 0
cv2.setTrackbarPos( first_track, orb_win, cur_first_lvl )
orb.setNLevels( cur_nlevels )
orb.setFirstLevel( cur_first_lvl )
orb_and_match()
def update_firstlevel( val ):
global cur_first_lvl
cur_first_lvl = val
if cur_first_lvl >= cur_nlevels:
cur_first_lvl = cur_nlevels - 1
cv2.setTrackbarPos( first_track, orb_win, cur_first_lvl )
orb.setFirstLevel( cur_first_lvl )
orb_and_match()
def update_nb_lines( val ):
global nb_lines
nb_lines = val
orb_and_match()
def update_id( val ):
global id_cube
id_cube = val
orb_and_match()
cv2.createTrackbar( "P.E.Size", orb_win, cur_patch_size, max_patch_size,
update_patch_size )
cv2.createTrackbar( "Fast Th", orb_win, cur_fast_thres, max_fast_thres,
update_fast_treshold )
cv2.createTrackbar( "Scale Fac", orb_win, int( 10*(cur_scale_factor - 1.0)),
max_scale_factor,
update_scale_factor )
cv2.createTrackbar( "N Levels", orb_win, cur_nlevels, max_nlevels,
update_nlevels )
first_track = cv2.createTrackbar( "First Lvl", orb_win, cur_first_lvl,
max_first_level, update_firstlevel )
cv2.createTrackbar( "Nb Lines", orb_win, nb_lines, max_nb_lines,
update_nb_lines )
cv2.createTrackbar( "ID cube", orb_win, id_cube, max_id_cube,
update_id )
cv2.imshow( orb_win, base_img )
orb_and_match()
cv2.waitKey(-1)