forked from andreiolariu/kaggle-driver-telematics
-
Notifications
You must be signed in to change notification settings - Fork 0
/
heading.py
201 lines (174 loc) · 5.58 KB
/
heading.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
import numpy as np
import matplotlib.pyplot as plt
from data_access import DataAccess
import util
def get_angle(angle_window):
angle_window = np.array(angle_window)
normalized = angle_window.T / np.maximum(np.linalg.norm(angle_window, axis=1), 0.1)
average = np.mean(normalized, axis=1)
angle = util.get_angle([1,0], [0,0], average)
if average[1] < 0:
angle = 360 - angle
return angle
def get_ride_heading(ride, variations=False, moving_average_window=3, stops=False, version=1):
'''
I don't know exactly what this does. I was drunk when I wrote it.
'''
ride = np.array(ride)
if moving_average_window:
ride = util.movingaverage(ride, moving_average_window)
ROLL_STEP = 3
ride2 = np.array(ride)
ride1 = np.roll(ride2, ROLL_STEP, axis=0)
l = len(ride)
ride0 = np.hstack((np.ones((l,1)), np.zeros((l,1))))
ride0 = ride0[ROLL_STEP:]
ride1 = ride1[ROLL_STEP:]
ride2 = ride2[ROLL_STEP:]
a1 = ride0
a2 = ride2 - ride1
distances1 = np.linalg.norm(a1, axis=1)
distances = np.linalg.norm(a2, axis=1)
x = (a1 * a2).sum(1) / np.maximum(distances1 * distances, 0.1)
y = np.sign(a2[:,1])
np.seterr(all='ignore')
angles = np.arccos(x) * 180 / np.pi
np.seterr(all='print')
angles[y<0] = 360 - angles[y<0]
angles[distances < 2] = np.nan
is_stopped = []
angle = []
angle_window = []
WINDOW_SIZE = 2
MIN_SPEED = 2
direction = []
for i, dist in enumerate(distances):
if dist > MIN_SPEED:
angle_window.append(a2[i])
angle_window = angle_window[-WINDOW_SIZE:]
if len(angle_window) < WINDOW_SIZE:
direction.append(np.nan)
else:
d = get_angle(angle_window[-WINDOW_SIZE/2:]) - get_angle(angle_window[:WINDOW_SIZE/2])
if d > 180:
d -= 360
if d < -180:
d += 360
direction.append(d)
else:
direction.append(np.nan)
if dist < MIN_SPEED or len(angle_window) < WINDOW_SIZE:
is_stopped.append(True)
else:
is_stopped.append(False)
angle.append(get_angle(angle_window))
windows = []
current_window = []
current_window_type = 0
for i in range(len(direction)):
if np.isnan(direction[i]):
current_window = []
windows.append(0)
continue
current_window.append(direction[i])
current_window = current_window[-4:]
t = np.mean(current_window)
if current_window_type == 0:
if np.abs(t) > 3:
current_window_type = np.sign(t)
else:
if np.sign(current_window[-1]) != current_window_type:
current_window_type = 0
windows.append(current_window_type)
windows = windows[2:] + [0, 0]
sw = True
while sw:
sw = False
for i in range(1, len(windows) - 1):
if windows[i] != windows[i-1] and windows[i] != windows[i+1]:
windows[i] = windows[i+1]
sw = True
for i in range(3, len(windows)):
if windows[i-3] != windows[i-2] and \
windows[i-2] == windows[i-1] and \
windows[i-1] != windows[i]:
windows[i-2] = windows[i-3]
windows[i-1] = windows[i]
sw = True
description = []
current_window_type = 'stop' if stops else 'straight'
new_type = current_window_type
current_window_length = 0
for i in range(1, len(windows)):
if stops:
if is_stopped[i]:
if current_window_type != 'stop':
new_type = 'stop'
if windows[i] == 0 and not is_stopped[i]:
if current_window_type != 'straight':
new_type = 'straight'
else:
if windows[i] == 0 or is_stopped[i]:
if current_window_type != 'straight':
new_type = 'straight'
if windows[i] == 1:
if current_window_type != 'left':
new_type = 'left'
if windows[i] == -1:
if current_window_type != 'right':
new_type = 'right'
if new_type == current_window_type:
current_window_length += 1
else:
if current_window_length:
description.append([
current_window_type,
current_window_length,
util.euclidian_distance(ride2[i-current_window_length], ride2[i]),
-np.sum(direction[i-current_window_length : i-1])
])
current_window_type = new_type
current_window_length = 0
i = 0
while i < len(description) - 1:
if description[i][0] in ['right', 'left'] and np.abs(description[i][3]) < 15:
description[i+1][2] += description[i][2]
description.pop(i)
else:
i += 1
i = 0
while i < len(description) - 1:
if description[i][0] == description[i+1][0]:
description[i+1][1] += description[i][1]
description[i+1][2] += description[i][2]
description[i+1][3] += description[i][3]
description.pop(i)
else:
i += 1
# convert to words
DIST_TH = [0, 10, 50, 100, 250, 500, 1500]
if version == 1:
TIME_TH = [0, 2, 8, 50]
ANGLE_TH = [0, 35, 70, 110, 145]
else:
TIME_TH = [0, 8, 50]
ANGLE_TH = [0, 10, 30, 55, 80, 110, 145]
mirrored = {'straight': 'straight', 'left': 'right', 'right': 'left', 'stop': 'stop'}
words_original = []
words_mirror = []
for row in description:
if row[0] == 'stop':
v = np.digitize([row[1]], TIME_TH)[0]
elif row[0] == 'straight':
v = np.digitize([row[2]], DIST_TH)[0]
else:
v = np.digitize([np.abs(row[3])], ANGLE_TH)[0]
word = '%s_%s' % (row[0], v)
words_original.append(word)
word = '%s_%s' % (mirrored[row[0]], v)
words_mirror.append(word)
if variations:
words_inverted = list(reversed(words_mirror))
words_mirror_inverted = list(reversed(words_original))
return [words_original, words_mirror, words_inverted, words_mirror_inverted]
return words_original