-
Notifications
You must be signed in to change notification settings - Fork 1
/
track.py
177 lines (142 loc) · 4.67 KB
/
track.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
from filterpy.kalman import KalmanFilter
import numpy as np
np.set_printoptions(precision=2, suppress=True)
class Box(object):
def __init__(self, x1, y1, x2, y2):
self.x1 = x1
self.y1 = y1
self.x2 = x2
self.y2 = y2
@classmethod
def from_z(cls, px, py, scale, ratio):
import math
w = math.sqrt(scale * ratio)
h = scale / w
x1 = px - w/2
x2 = px + w/2
y1 = py - h/2
y2 = py + h/2
box = Box(x1, y1, x2, y2)
return box
def get_bb(self):
return int(self.x1), int(self.y1), int(self.x2), int(self.y2)
def get_z(self):
z = np.array([self._px(), self._py(), self._scale(), self._ratio()]).reshape(-1,1)
return z
def _px(self):
px = (self.x1 + self.x2) / 2
return px
def _py(self):
py = (self.y1 + self.y2) / 2
return py
def _scale(self):
w = self.x2 - self.x1
h = self.y2 - self.y1
scale = w*h
return scale
def _ratio(self):
w = self.x2 - self.x1
h = self.y2 - self.y1
ratio = float(w)/h
return ratio
RELIABLE_THD = 20
DRAW_THD = 5
UNTRACK_THD = 5
class BoxTracker(object):
_N_MEAS = 4 # (px, py, scale, ratio)-ordered
_N_STATE = 7 # (px, py, scale, ratio, vx, vy, vs)-ordered
def __init__(self, init_box, group_number=1):
self._kf = self._build_kf(init_box)
self.group_number = group_number
self.detect_count = 1
self.miss_count = 0
def _build_kf(self, init_box, Q_scale=1.0, R_scale=400.0):
kf = KalmanFilter(dim_x=self._N_STATE,
dim_z=self._N_MEAS)
kf.F = np.array([[1,0,0,0,1,0,0],
[0,1,0,0,0,1,0],
[0,0,1,0,0,0,1],
[0,0,0,1,0,0,0],
[0,0,0,0,1,0,0],
[0,0,0,0,0,1,0],
[0,0,0,0,0,0,1]])
kf.H = np.array([[1,0,0,0,0,0,0],
[0,1,0,0,0,0,0],
[0,0,1,0,0,0,0],
[0,0,0,1,0,0,0]])
Q = np.zeros_like(kf.F)
for i in range(self._N_MEAS, self._N_STATE):
Q[i, i] = Q_scale
R = np.eye(self._N_MEAS) * R_scale
kf.Q = Q
kf.R = R
kf.x = np.zeros((self._N_STATE, 1))
kf.x[:self._N_MEAS,:] = init_box.get_z()
return kf
def predict(self):
if self.is_missing_but_drawing():
# v_scale
self._kf.x[6,0] = 0
self._kf.predict()
predict_box = Box.from_z(*self._kf.x[:self._N_MEAS,0])
return predict_box
def update(self, box=None):
"""
# Args
box : Box instance
in the case of no matching box, box argument is None
# Returns
filtered_box : Box instance
"""
if box is not None:
self._detect_counting()
z = box.get_z()
self._kf.update(z)
filtered_box = Box.from_z(*self._kf.x[:self._N_MEAS,0])
return filtered_box
def miss(self):
self.miss_count += 1
def get_bb(self):
box = Box.from_z(*self._kf.x[:4,0])
bounding_box = box.get_bb()
return bounding_box
def is_draw(self):
if self._is_reliable_target() or self.detect_count-self.miss_count >= DRAW_THD:
return True
else:
return False
def is_delete(self):
def _in_reliable_range():
x1, y1, x2, y2 = self.get_bb()
margin = 50
# hard coding
if x1 > margin and x2 < 1280-margin and y1 > 350 and y2 < 960-margin:
return True
else:
return False
if self.is_missing_but_drawing():
if _in_reliable_range():
return False
else:
return True
else:
if self.miss_count > self.detect_count or self.miss_count >= UNTRACK_THD:
return True
else:
return False
def is_missing_but_drawing(self):
if self._is_reliable_target() and self.miss_count >= UNTRACK_THD:
return True
else:
return False
def _is_reliable_target(self):
if self.detect_count >= RELIABLE_THD:
return True
else:
return False
def _detect_counting(self):
self.detect_count += 1
if self._is_reliable_target():
self.miss_count = 0
if __name__ == "__main__":
pass