-
Notifications
You must be signed in to change notification settings - Fork 10
/
Lane_Detection.py
492 lines (373 loc) · 14.2 KB
/
Lane_Detection.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
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
import numpy as np
import cv2 as cv
from skimage import feature
from collections import defaultdict
from skimage.transform import hough_line, hough_line_peaks
import math as mt
import operator
from skimage import io
from GenerateBEV import BevParams
from GenerateBEV import Calibration
from GenerateBEV import BirdsEyeView
np.set_printoptions(threshold=np .nan)
#debug = True
debug = False
class Line:
startpoint = []
endpoint = []
linescore = 0
list_x_y_points = []
def __init__(self, startpoint, endpoint, linescore):
self.startpoint = startpoint
self.endpoint = endpoint
self.linescore = linescore
def FilterLines(input_image, width_kernel_x,width_kernel_y,sigmax,sigmay):
#Changing Data Type
temp_image = np.float32(input_image)
#Scaling
temp_image = temp_image*1./255
# if(debug):
# cv.imshow("Scaled_Down_Image", temp_image)
# cv.waitKey(0)
#Computer x and y kernels for Filtering Image
#y - Gaussian Kernel
#x - Derivative of Gaussian
g_kernel_y = []
variance_y = sigmay*sigmay
for i in range(-width_kernel_y,width_kernel_y+ 1):
k1 = np.exp((-0.5/variance_y)*i*i)
g_kernel_y.append(k1)
g_kernel_y = np.array(g_kernel_y, dtype = np.float32,ndmin=2)
g_kernel_y = np.reshape(g_kernel_y,(2*width_kernel_y+1,1))
variance = sigmax*sigmax
g_kernel_x = []
for i in range(-width_kernel_x,width_kernel_x+1 ):
k1 = np.exp(-i*i*0.5/variance)
function = (1/variance)*k1 - (i*i)/(variance*variance)*k1
g_kernel_x.append(function)
g_kernel_x = np.array(g_kernel_x, dtype = np.float32,ndmin=2)
kernel = g_kernel_y.dot(g_kernel_x)
mean = cv.mean(kernel)[0]
kernel = cv.subtract(kernel,mean)
temp_image = cv.filter2D(temp_image,-1,kernel)
return temp_image
#Get Specified quantile value from Input image
def getRansacLines(thresholded_image, lines):
e1 = cv.getTickCount()
test_image = np.transpose(thresholded_image)
non_zero_points = np.nonzero(test_image)
list_x_y_points = []
for i in range(0, len(non_zero_points[0])):
point = (non_zero_points[0][i], non_zero_points[1][i])
list_x_y_points.append(point)
sorted_by_second = sorted(list_x_y_points, key=lambda tup: tup[1])
# print(sorted_by_second)
intializepointsinROI(sorted_by_second, lines)
for i in range(0, len(lines)):
# print(lines[i].list_x_y_points)
data = np.array(lines[i].list_x_y_points)
line = cv.fitLine(data,cv.DIST_FAIR,0,0.01,0.01)
mult = max(gray_image.shape[0], gray_image.shape[1])
startpoint = (int(line[2] - mult*line[0]), int(line[3] - mult*line[1]))
endpoint = (int(line[2] + mult*line[0]), int(line[3] + mult*line[1]))
points = cv.clipLine((0,0,gray_image.shape[1],gray_image.shape[0]), startpoint, endpoint)
x_limit_max = max(lines[i].startpoint[0], lines[i].endpoint[0])
x_limit_min = min(lines[i].startpoint[0], lines[i].endpoint[0])
points = [list(i) for i in points[1:]]
# print(points[][0])
for i in range(0, len(points)):
if(points[i][0] < x_limit_min ):
points[i][0] = x_limit_min
elif(points[i][0] > x_limit_max):
points[i][0] = x_limit_max
# print(points)
cv.line(gray_image, tuple(points[0]), tuple(points[1]),(0, 0, 255),2)
# # print(line)
#cv.imshow("Result", gray_image)
#cv.waitKey(0)
#Write Image
e2 = cv.getTickCount()
time = (e2 -e1)/cv.getTickFrequency()
print("Time for Fitting line")
print(time)
#cv.imshow("Result", gray_image)
#cv.waitKey(0)
#cv.imwrite("/home/nvidia/test_image.png", gray_image)
def intializepointsinROI(x_y_points, lines):
# for i in range(0, len(lines)):
# print(lines[i].startpoint)
# print(lines[i].endpoint)
# threshold = x_limit_max
for i in range(0, len(lines)):
# print(lines[i].startpoint)
# print(lines[i].endpoint)
x_limit_max = max(lines[i].startpoint[0], lines[i].endpoint[0])
x_limit_min = min(lines[i].startpoint[0], lines[i].endpoint[0])
# threshold = x_limit_max - x_limit_min
# print(threshold)
search_range = np.arange(x_limit_min-1, x_limit_max+1)
# print(search_rangenge)
points = [x for x in x_y_points if x[0] in search_range]
lines[i].list_x_y_points = points
# print(lines[i].list_x_y_points)
#The fucntion gets the specified quantile value
#from the input image
#Param:-
#input_image - Filtered_Image
#qtile - Quantile Threshold
#
def getQuantile(input_image, qtile):
number_rows = input_image.shape[0]
number_columns = input_image.shape[1]
temp_image = np.reshape(input_image,(1,number_rows*number_columns))
quantile = getPoints(temp_image,qtile)
output_image = thresholdlower(input_image,quantile)
return output_image[1]
def getPoints(input_image, quantile):
# print(input_image.size)
size = input_image.size
if (size == 0):
return float(0)
elif (size ==1 ):
return input_image[0]
elif (quantile<=0):
return np.amin(input_image)
elif (quantile>=1):
return np.amax(input_image)
pos = (size-1)*quantile
index = np.uint32(pos)
delta = pos - index
temp_image = input_image
temp_image_1 = np.partition(temp_image,index)
i1 = temp_image_1[[0],index]
i2 = np.amin(temp_image_1[[0],index +1:])
return (i1*(1.0 - delta) + i2*delta)
#Threshold Image
def thresholdlower(input_image, threshold):
# print(input_image)
output_image = cv.threshold(input_image,threshold,0,cv.THRESH_TOZERO)
# cv.imshow("Result",output_image[1])
# cv.waitKey(0)
return output_image
def getbinaryimage(input_image):
# Binarize the input image
#Calculate Maximum and minimum of an image
maximum = np.amax(input_image)
# print(maximum)
minimum = np.amin(input_image)
thresh = (maximum - minimum)/2
output_image = cv.threshold(input_image,thresh,1,cv.THRESH_BINARY)
return output_image[1]
def getclearImage(thresholded_image):
img_shape = thresholded_image.shape
ncolumns = img_shape[1]
nrows = img_shape[0]
approx_1 = int(nrows*0.75)
approx = int(ncolumns*0.75)
img_copy = thresholded_image
#Make Right End zero
img_copy[:,approx:] = 0
img_copy[:,0:(ncolumns-approx)] = 0
img_copy[0:(nrows-approx_1),:] = 0
return img_copy
def getHoughLines(input_image):
# cv.imshow("Intermediate",input_image)
# cv.waitKey(0)
e1 = cv.getTickCount()
hspace, angles, dist = hough_line(binary_image)
maximum = np.amax(hspace)
# print(maximum)
peak_hspace, angles, dist = hough_line_peaks(hspace, angles,dist, threshold = maximum*0.38,min_distance = 20)
# print(peak_hspace)
if(debug):
for i in range(0, len(angles)):
a = np.cos(angles[i])
b = np.sin(angles[i])
x0 = a*dist[i]
y0 = b*dist[i]
x1 = int(x0 + binary_image.shape[0]*(-b))
y1 = int(y0 + binary_image.shape[0]*(a))
x2 = int(x0 - binary_image.shape[0]*(-b))
y2 = int(y0 - binary_image.shape[0]*(a))
# print(x1)
# print(y1)
cv.line(gray_image,(x1,y1),(x2,y2),(0,0,255),1)
io.imsave("/home/mohak/Process_Pipeline/Initial_Guess_For_Ransac.png",gray_image)
cv.imshow("Result", gray_image)
cv.waitKey(0)
# print(peak_hspace)
lines = groupLines(angles,dist,peak_hspace)
# print(lines)
lines = checklanewidth(lines)
e2 = cv.getTickCount()
time = (e2 - e1)/cv.getTickFrequency()
print("Time for Initial Approximations")
print(time)
#print(lines)
getRansacLines(input_image,lines)
return hspace, angles, dist
def checklanewidth(lines):
number_of_lanes = len(lines)
#Sort Line Objects According to Starting Point
lines.sort(key = lambda x:x.startpoint[0])
# print(lines)
min_distance = 10 #approx
max_distance_two_side_lanes = 45 #approx
max_distance_two_edge_lanes = 70 #approx
x_points = []
for i in range(0, number_of_lanes):
x_max = max(lines[i].startpoint[0], lines[i].endpoint[0])
x_points.append(x_max)
x_points.sort()
# print(x_points)
if (number_of_lanes == 2):
#Calculate Diff array
diff_array = np.diff(x_points)
for i in range(0, len(diff_array)):
if(diff_array[i] < min_distance or diff_array[i] > max_distance_two_edge_lanes):
lines.pop(i+1)
elif(number_of_lanes == 3):
diff_array = np.diff(x_points)
for i in range(0, len(diff_array)):
if (diff_array[i] < min_distance or diff_array[i] > max_distance_two_side_lanes):
lines.pop(i+1)
return lines
def getlocalMaxima(input_matrix, threhold):
rows = input_matrix.shape[0]
columns = input_matrix.shape[1]
localMaxima = []
localMaximaLoc = []
localMaximaLoc = feature.peak_local_max(input_matrix,min_distance=3,indices=True)
for i,j in localMaximaLoc:
localMaxima.append(input_matrix[i][j])
data_dict = defaultdict(list)
for i in range(0,len(localMaxima)):
data_dict[localMaxima[i]].append(localMaximaLoc[i])
data_dict = sorted(data_dict.items(),reverse = True)
return data_dict
def getLineEndPoints(r, theta,img_size):
startpoint = []
endpoint = []
if(mt.cos(theta) == 0):
xup = int(img_size[0])
xdown = int(img_size[0])
else:
xup = int(r/mt.cos(theta))
xdown = int((r-img_size[1]*mt.sin(theta))/mt.cos(theta))
if(mt.sin(theta ) ==0):
yleft = int(img_size[1])
yright = int(img_size[1])
else:
yleft = int(r/mt.sin(theta))
yright = int((r-img_size[0]*mt.cos(theta))/mt.sin(theta))
pts = [(xup,0),(xdown,img_size[1]),(0,yleft),(img_size[0],yright)]
count = 0
for i in range(0,4):
if(isPointInside(pts[i],img_size)):
startpoint.append((pts[i][0],pts[i][1]))
count = i
break
# print(count )
for i in range(count+1,4):
if(isPointInside(pts[i],img_size)):
endpoint.append((pts[i][0],pts[i][1]))
break
return startpoint, endpoint
def isPointInside(point, img_size):
if(point[0] >= 0 and point[0] <= img_size[0] and point[1] >= 0 and point[1] <= img_size[1]):
return True
else:
return False
def groupLines(angles, dist, peak_hspace):
startpoints = []
endpoints = []
for i in range(0, len(angles)):
startpoint, endpoint = getLineEndPoints(dist[i], angles[i], (binary_image.shape[1],binary_image.shape[0]))
startpoints.append(startpoint)
endpoints.append(endpoint)
number_of_lines = len(dist)
lines = []
# print(startpoints[0][0])
for i in range(0, number_of_lines):
line = Line(startpoints[i][0],endpoints[i][0], peak_hspace[0])
lines.append(line)
# print(lines)
return lines
#Take Input Test Image
#Generate IPM view in same file
input_image = cv.imread("/home/nvidia/Lane_Detection/Original_Images/img_1.png",cv.IMREAD_COLOR)
e1 = cv.getTickCount()
#Create BEV object
bev = BirdsEyeView()
#Camera Extrinsic Parameters
P2 = np.array([[7.070912000000e+02,0.000000000000e+00,6.018873000000e+02,4.688783000000e+01],
[ 0.000000000000e+00, 7.070912000000e+02, 1.831104000000e+02, 1.178601000000e-01],
[0.000000000000e+00, 0.000000000000e+00, 1.000000000000e+00, 6.203223000000e-03]],dtype=np.float64)
R0_rect = np.array([[9.999280000000e-01,8.085985000000e-03,-8.866797000000e-03],
[-8.123205000000e-03, 9.999583000000e-01,-4.169750000000e-03],
[8.832711000000e-03,4.241477000000e-03,9.999520000000e-01]], dtype = np.float64)
Tr_cam_to_road = np.array([[9.998675805558e-01, -1.466259288355e-02, -7.059878200710e-03, 2.879062998184e-02],
[1.469236542096e-02, 9.998832652489e-01, 4.183808189280e-03, -1.630891383620e+00],
[6.997709379545e-03, -4.286980067163e-03, 9.999662905842e-01,3.368200142169e-01]], dtype =np.float64)
bev.setup(P2,R0_rect,Tr_cam_to_road)
output_image = bev.compute(input_image)
e2 = cv.getTickCount()
time = (e2-e1)/cv.getTickFrequency()
print("Time for IPM")
print(time)
#input_image = cv.imread("/home/nvidia/Lane_Detection/Test_Images/IPM_test_image_4.png")
input_image = output_image
#Input IPM image for Lane Detection
#input_image = output_image
#cv.imshow("Input_Image", input_image)
#cv.waitKey(0)
e5 = cv.getTickCount()
#input_image = cv.imread("/home/nvidia/Lane_Detection/Test_Images/IPM_test_image_1.png")
if(debug):
cv.imshow("Input_Image", input_image)
cv.waitKey(0)
#GrayScale Image
gray_image = cv.cvtColor(input_image,cv.COLOR_BGR2GRAY)
if(debug):
cv.imshow("Gray_Image", gray_image)
cv.waitKey(0)
#Preprocess Image
filtered_image = FilterLines(gray_image,2,2,2,10)
if(debug):
cv.imshow("Gray_Image", filtered_image)
cv.waitKey(0)
io.imsave("/home/mohak/Process_Pipeline/filtered_image.png",filtered_image)
#Threshold Image
thresholded_image = getQuantile(filtered_image,0.985)
if(debug):
io.imsave("/home/mohak/Process_Pipeline/thresholded_image.png",thresholded_image)
cv.imshow("Gray_Image", thresholded_image)
cv.waitKey(0)
# Clean Negetive parts of the Image
thresholded_image = thresholdlower(thresholded_image,0)[1]
# print(thresholded_image.shape)
#Clean Outer Edges of the Images
clear_image = getclearImage(thresholded_image)
if(debug):
cv.imshow("Cleaned_Image", thresholded_image[:,150:])
cv.imshow("Cleaned_Image", clear_image)
cv.waitKey(0)
# Binarize the Image
binary_image = getbinaryimage(thresholded_image)
if(debug):
cv.imshow("Result", binary_image)
cv.waitKey(0)
#Select ROI
image_height = binary_image.shape[0]
image_width = binary_image.shape[1]
ROI_height = int(0.45*image_height)
ROI_image = binary_image[ROI_height:,:]
if(debug):
io.imsave("/home/mohak/Process_Pipeline/binary_image_after_ROI.png", ROI_image)
cv.imshow("Result", ROI_image)
cv.waitKey(0)
e6 = cv.getTickCount()
time = (e6-e5)/cv.getTickFrequency()
print("Pre-processing Time")
print(time)
hspace, angles, dist = getHoughLines(ROI_image)