-
Notifications
You must be signed in to change notification settings - Fork 1
/
LK_Affine_Tracker.py
237 lines (194 loc) · 7 KB
/
LK_Affine_Tracker.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
import glob
import cv2 as cv
import numpy as np
def get_coordinates_array(x_range,y_range):
a=(x_range[1]-x_range[0]) + 1
b=(y_range[1]-y_range[0]) + 1
coordinates=np.zeros((3,a*b))
count=0
for y in range(y_range[0],y_range[1]+1):
for x in range(x_range[0],x_range[1]+1):
coordinates[0,count]=x
coordinates[1,count]=y
coordinates[2,count]=1
count+=1
return coordinates
def point_selector(T_x_image):
points = cv.selectROI(T_x_image)
x_start = points[0]
x_end = points[0]+points[2]
y_start = points[1]
y_end = points[1]+ points[3]
x_range=[x_start,x_end]
y_range=[y_start,y_end]
return x_range,y_range
def get_new_coord(T_x_coordinates,p,x_range,y_range):
x1,x2=x_range
y1,y2=y_range
vertex_array=np.array([[x1,x1,x2,x2],[y1,y2,y2,y1],[1,1,1,1]])
affine_mat =np.zeros((2,3))
count =0
for i in range(3):
for j in range(2):
affine_mat[j,i]= p[count,0]
count =count+1
affine_mat[0,0]+=1
affine_mat[1,1]+=1
new_vertex_array=np.dot(affine_mat,vertex_array)
new_coordinates=np.dot(affine_mat,T_x_coordinates)
new_coordinates = new_coordinates.astype(int)
return new_coordinates,new_vertex_array
def get_pixel_array(image,coordinates):
img_array = np.zeros((1,coordinates.shape[1]))
img_array[0,:]=image[coordinates[1,:],coordinates[0,:]]
return img_array
def compute_error(img_array1,img_array2):
error=img_array1-img_array2
return error
def get_T_x_array(Template_image, x_range, y_range):
T_x_coordinates=get_coordinates_array(x_range,y_range)
p_T_x=np.array([[0,0,0,0,0,0]]).T
new_coordinates,new_ver=get_new_coord(T_x_coordinates,p_T_x,x_range,y_range)
T_x_img_arr = get_pixel_array(Template_image,new_coordinates)
return T_x_img_arr
def convert_lab(image):
clahe = cv.createCLAHE(clipLimit=1., tileGridSize=(1,1))
lab = cv.cvtColor(image, cv.COLOR_BGR2LAB)
l, a, b = cv.split(lab)
l2 = clahe.apply(b)
lab = cv.merge((l,a,l2))
img2 = cv.cvtColor(lab, cv.COLOR_LAB2BGR)
return img2
def array_to_img(arrayimage,x_range,y_range):
a=(x_range[1]-x_range[0]) + 1
b=(y_range[1]-y_range[0]) + 1
img=arrayimage.astype(np.uint8)
image=np.reshape(img,(b,a))
return image
def sobel(img):
# sobelx = cv.Scharr(img,cv.CV_64F,1,0)
# sobely = cv.Scharr(img,cv.CV_64F,0,1)
sobelx = cv.Sobel(img,cv.CV_64F,1,0,ksize=3)
sobely = cv.Sobel(img,cv.CV_64F,0,1,ksize=3)
return sobelx,sobely
def jacobian(x,y):
mat=np.array(([x,0,y,0,1,0],[0,x,0,y,0,1]))
return mat
def get_delta_p(err_array,steep_descent):
SDParam = np.dot(steep_descent.T, err_array.T)
Hessian = np.dot(steep_descent.T, steep_descent)
Hessian_Inv = np.linalg.pinv(Hessian)
delta_p = np.dot(Hessian_Inv, SDParam)
return delta_p
def update_p (p,delta_p):
p = np.reshape(p,(6,1))
p = p+delta_p
return p
def sanitycheck(coorinate_array, img) :
min_xy=np.amin(coorinate_array,axis=1)
max_xy=np.amax(coorinate_array,axis=1)
if min_xy[0] < 0 or max_xy[0] >= img.shape[1] or min_xy[1] < 0 or max_xy[1] >= img.shape[0]:
T = False
else :
T = True
return T
def draw_rectangle(image,pts):
cv.polylines(image, np.int32([pts.T]), 1, (0, 255, 0), 2)
return image
def get_norm(del_p):
return np.linalg.norm(del_p)
def get_steep_descent(sobelx,sobely,new_coordinates,old_coordinates):
sobel_delx_array=get_pixel_array(sobelx,new_coordinates)
sobel_dely_array=get_pixel_array(sobely,new_coordinates)
img1=sobel_delx_array*old_coordinates[0,:]
img2=sobel_dely_array*old_coordinates[0,:]
img3=sobel_delx_array*old_coordinates[1,:]
img4=sobel_dely_array*old_coordinates[1,:]
steepest_descent_image=np.vstack((img1,img2,img3,img4, sobel_delx_array, sobel_dely_array)).T
return steepest_descent_image
def affineLKtracker(T_x_coordinates,T_x_array,gray_image,x_range, y_range, p, sobelx, sobely):
new_img_coordinates,new_vertex_array=get_new_coord(T_x_coordinates,p,x_range,y_range)
if(sanitycheck(new_img_coordinates,gray_image)):
sanity =True
img_arr = get_pixel_array(gray_image,new_img_coordinates)
error_array = compute_error(T_x_array,img_arr)
steep_descent=get_steep_descent(sobelx,sobely,new_img_coordinates,T_x_coordinates)
del_p=get_delta_p(error_array,steep_descent)
p_norm = get_norm(del_p)
p = update_p(p,del_p)
else:
sanity =False
del_p = np.array([[0,0,0,0,0,0]]).T
p_norm = get_norm(del_p)
return p, del_p, p_norm,new_img_coordinates,new_vertex_array,sanity
# Starting
print("\n\nPlease enter 0 for Car, 1 for human, 2 for vase : ")
choice = int(input())
#threshold = 0.01
if choice ==0 :
path = "data/car/*.jpg"
path1 = "output/car/car"
threshold = 0.03
if choice ==1 :
path = "data/human/*.jpg"
path1 = "output/human/human"
threshold = 0.05
if choice ==2 :
path = "data/vase/*.jpg"
path1 = "output/vase/vase"
threshold = 0.005
filenames = [img for img in glob.glob(path)]
filenames.sort() # ADD THIS LINE
T_x_image=cv.imread(filenames[0])
T_x_image=convert_lab(T_x_image)
T_x_image=cv.cvtColor(T_x_image,cv.COLOR_BGR2GRAY)
t_x_mean=np.mean(T_x_image)
if choice==0 :
print("Enter 0 for best result, enter 1 to choose your own ROI ")
ch = int(input())
if ch==0:
x_range=[136,338]
y_range=[108,275]
elif ch==1:
x_range,y_range=point_selector(T_x_image)
if choice==1 :
print("Enter 0 for best result, enter 1 to choose your own ROI ")
ch = int(input())
if ch==0:
x_range=[264,283]
y_range=[292,360]
elif ch==1:
x_range,y_range=point_selector(T_x_image)
if choice==2 :
print("Enter 0 for best result, enter 1 to choose your own ROI ")
ch = int(input())
if ch==0:
x_range=[127,172]
y_range=[92,147]
elif ch==1:
x_range,y_range=point_selector(T_x_image)
print(x_range,y_range)
#x_range=[264,283] ### for Human best result
#y_range=[292,360] ### for Human best result
T_x_array = get_T_x_array(T_x_image,x_range,y_range)
T_x_coordinates=get_coordinates_array(x_range,y_range)
p=np.array([[0,0,0,0,0,0]]).T
count = 0
for img in filenames:
# print(count)
image=cv.imread(img)
image=convert_lab(image)
cv.waitKey(1)
gray=cv.cvtColor(image,cv.COLOR_BGR2GRAY)
i_x_mean=np.mean(gray)
gray=(gray*((t_x_mean/i_x_mean))).astype(float)
sobelx,sobely = sobel(gray)
count = count+1
while True:
p, del_p, p_norm,new_img_coordinates,new_vertex,sane= affineLKtracker(T_x_coordinates,T_x_array,gray,x_range,y_range,p,sobelx,sobely)
if p_norm <= threshold or sane == False :
break
rect_img = draw_rectangle(image, new_vertex)
cv.imshow('rect',rect_img)
cv.imwrite(path1+'%s.jpg'%count,rect_img)
cv.destroyAllWindows()