-
Notifications
You must be signed in to change notification settings - Fork 0
/
lane_detector.py
152 lines (114 loc) · 5.58 KB
/
lane_detector.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
import cv2
import numpy as np
import os
from buffered_poly_fit import BufferedPolyFit
from buffered_number import BufferedNumber
class LaneDetector:
def __init__(self, camera, perspective, color_mask, fit_sample_size=5000, y_check_steps=15, max_x_std=50, pixels_per_meter=300):
self.camera = camera
self.perspective = perspective
self.left_poly = BufferedPolyFit(2)
self.right_poly = BufferedPolyFit(2)
self.fit_sample_size = fit_sample_size
self.color_mask = color_mask
self.y_check_steps = y_check_steps
self.max_x_std = max_x_std
self.pixels_per_meter = pixels_per_meter
self.curvature_buffer = BufferedNumber()
self.lane_position_buffer = BufferedNumber()
def apply_color_mask(self, img):
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
mask = np.zeros(img.shape[:2], np.uint8)
for color_range in self.color_mask:
mask = cv2.bitwise_or(mask, cv2.inRange(hsv, color_range[0], color_range[1]))
return cv2.bitwise_and(gray, mask)
def get_lane_points(self, img, offset=0):
y, x = np.nonzero(img)
# Sample fewer points without replacement to improve processing speed
i = np.random.choice(len(x), self.fit_sample_size)
y = y[i]
x = x[i]
# Keep points where deviation along x-axis is not too big
keep = [np.array([], np.uint8), np.array([], np.uint8)]
discard = [np.array([], np.uint8), np.array([], np.uint8)]
step = img.shape[0] // self.y_check_steps
for row in range(0, img.shape[0], step):
in_row = np.where((y >= row) & (y <= row + step))
if not len(in_row):
continue
if np.std(x[in_row]) < self.max_x_std:
keep[0] = np.concatenate((keep[0], y[in_row]))
keep[1] = np.concatenate((keep[1], x[in_row]))
else:
discard[0] = np.concatenate((discard[0], y[in_row]))
discard[1] = np.concatenate((discard[1], x[in_row]))
keep[1] += offset
discard[1] += offset
return keep, discard
def fit_lanes(self, img):
#kernel = cv2.getStructuringElement(shape=cv2.MORPH_RECT, ksize=(5, 5))
#img = cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel, iterations=1)
output = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
left, left_discarded = self.get_lane_points(img[:,0:img.shape[1]//3])
right, right_discarded = self.get_lane_points(img[:,img.shape[1]//3*2:], img.shape[1]//3*2)
left_line = self.left_poly.fit(left[0], left[1])
right_line = self.right_poly.fit(right[0], right[1])
# Draw lane
if left_line and right_line:
for y in range(img.shape[0]):
l, r = int(left_line(y)), int(right_line(y))
cv2.line(output, (l, y), (r, y), (0, 255, 0), 1)
for points in (left, right):
for i in range(len(points[0])):
cv2.circle(output, (points[1][i], points[0][i]), 1, (255, 0, 0), 1)
for points in (left_discarded, right_discarded):
for i in range(len(points[0])):
cv2.circle(output, (points[1][i], points[0][i]), 1, (0, 0, 255), 1)
return output
def run(self, raw, save_images=False):
if save_images:
try:
os.makedirs(save_images)
except FileExistsError:
pass
cv2.imwrite("%s/1-raw.jpg" % save_images, raw)
# Undistort
img = self.camera.undistort(raw)
if save_images:
cv2.imwrite("%s/2-undistorted.jpg" % save_images, img)
# Bird's eye view
img = self.perspective.transform(img)
if save_images:
cv2.imwrite("%s/3-birds-eye-view.jpg" % save_images, img)
# Apply lane color mask
img = self.apply_color_mask(img)
if save_images:
cv2.imwrite("%s/4-color-mask.jpg" % save_images, img)
# Fit lanes
img = self.fit_lanes(img)
if save_images:
cv2.imwrite("%s/6-fit.jpg" % save_images, img)
# Transform lanes onto original image
img = cv2.addWeighted(raw, 1, self.camera.redistort(self.perspective.invert(img)), 0.5, 0)
# Write curvature and position information
if self.left_poly.last_fit and self.right_poly.last_fit:
curvature = (self.left_poly.get_curvature(self.camera.height) +\
self.left_poly.get_curvature(self.camera.height)) / self.pixels_per_meter
curvature = self.curvature_buffer.get(curvature)
if abs(curvature) > 20:
turn = "otse"
elif curvature < 0:
turn = "vasakule"
else:
turn = "paremale"
lane_center = self.right_poly.last_fit(self.camera.height) + self.left_poly.last_fit(self.camera.height) / 2
lane_position = (self.camera.width / 2 - lane_center) / self.pixels_per_meter
lane_position = self.lane_position_buffer.get(lane_position)
cv2.putText(img, "Kurvi raadius: %.2f m (%s)" % (curvature, turn), (10, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 1)
cv2.putText(img, "Asukoht keskjoone suhtes: %.2f m" % lane_position, (10, 40),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 1)
if save_images:
cv2.imwrite("%s/7-final.jpg" % save_images, img)
return img