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
| """
基于3D人体姿态估计的OOP检测
输入: DMS/OMS摄像头 (RGB-IR, 全局快门)
模型: 轻量化3D HPE (Human Pose Estimation)
输出: 关键点3D坐标 → 姿态分类
关键点定义(COCO-17 + 车内扩展):
0: 鼻子 1-4: 眼睛耳朵 5-6: 肩膀
7-8: 手肘 9-10: 手腕 11-12: 髋部
13-14: 膝盖 15-16: 脚踝
扩展: 17-18: 头顶 19-20: 安全带位置
"""
import numpy as np
from typing import List
KEYPOINT_NAMES = [
"nose", "left_eye", "right_eye", "left_ear", "right_ear",
"left_shoulder", "right_shoulder", "left_elbow", "right_elbow",
"left_wrist", "right_wrist", "left_hip", "right_hip",
"left_knee", "right_knee", "left_ankle", "right_ankle"
]
class OOPDetector:
"""OOP异常姿态检测器"""
def __init__(self):
self.normal_sitting = {
'torso_angle': 90,
'head_angle': 0,
'shoulder_symmetry': 0,
'hip_position': 'seat',
}
def classify_posture(self, keypoints_3d: np.ndarray) -> dict:
"""
基于3D关键点分类乘员姿态
Args:
keypoints_3d: shape=(17, 3), 3D坐标(x,y,z) 单位cm
Returns:
posture classification
"""
shoulder_mid = (keypoints_3d[5] + keypoints_3d[6]) / 2
hip_mid = (keypoints_3d[11] + keypoints_3d[12]) / 2
torso_vec = shoulder_mid - hip_mid
torso_angle = np.degrees(np.arctan2(torso_vec[1], np.sqrt(torso_vec[0]**2 + torso_vec[2]**2)))
head_vec = keypoints_3d[0] - shoulder_mid
head_yaw = np.degrees(np.arctan2(head_vec[0], head_vec[2]))
head_pitch = np.degrees(np.arctan2(head_vec[1], np.sqrt(head_vec[0]**2 + head_vec[2]**2)))
shoulder_z_diff = abs(keypoints_3d[5][2] - keypoints_3d[6][2])
knee_mid = (keypoints_3d[13] + keypoints_3d[14]) / 2
leg_vec = knee_mid - hip_mid
leg_angle = np.degrees(np.arctan2(leg_vec[1], np.sqrt(leg_vec[0]**2 + leg_vec[2]**2)))
posture = "normal"
risk_level = 0
if torso_angle < 50:
posture = "reclined_excessive"
risk_level = 3
elif torso_angle < 65:
posture = "reclined_moderate"
risk_level = 2
if abs(head_yaw) > 60:
posture = "looking_sideways" if posture == "normal" else posture + "+sideways"
risk_level = max(risk_level, 2)
if abs(head_yaw) > 120:
posture = "rear_facing"
risk_level = 3
if torso_angle > 110:
posture = "leaning_forward"
risk_level = 3
if leg_angle > 80 and keypoints_3d[15][1] > keypoints_3d[11][1]:
posture = "feet_on_dash" if posture == "normal" else posture + "+feet_up"
risk_level = max(risk_level, 2)
return {
'posture': posture,
'risk_level': risk_level,
'torso_angle': round(float(torso_angle), 1),
'head_yaw': round(float(head_yaw), 1),
'head_pitch': round(float(head_pitch), 1),
'shoulder_z_diff': round(float(shoulder_z_diff), 2),
'leg_angle': round(float(leg_angle), 1),
'submarining_risk': self.estimate_submarining(torso_angle)
}
def estimate_submarining(self, torso_angle: float) -> str:
"""基于躯干角度估计submarining风险"""
if torso_angle >= 90:
return "低"
elif torso_angle >= 65:
return "中"
elif torso_angle >= 50:
return "高"
else:
return "极高"
if __name__ == "__main__":
det = OOPDetector()
normal_kp = np.array([
[0, 50, 30], [-2, 52, 30], [2, 52, 30],
[-5, 50, 30], [5, 50, 30],
[0, 45, 25], [0, 45, 25],
[0, 35, 25], [0, 35, 25],
[0, 25, 25], [0, 25, 25],
[0, 20, 25], [0, 20, 25],
[0, 10, 28], [0, 10, 28],
[0, 0, 30], [0, 0, 30],
], dtype=float)
r1 = det.classify_posture(normal_kp)
print(f"正常坐姿: {r1['posture']} (风险={r1['risk_level']}, 躯干={r1['torso_angle']}°)")
print(f" Submarining风险: {r1['submarining_risk']}")
reclined_kp = normal_kp.copy()
reclined_kp[5:7, 1] -= 15
reclined_kp[5:7, 2] += 20
reclined_kp[11:13, 1] -= 5
r2 = det.classify_posture(reclined_kp)
print(f"\n后靠坐姿: {r2['posture']} (风险={r2['risk_level']}, 躯干={r2['torso_angle']}°)")
print(f" Submarining风险: {r2['submarining_risk']}")
|
