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
| """
Aptiv AOC 摄像头乘员分类模型架构
核心技术：
1. 目标检测：YOLO/MediaPipe 人体关键点
2. 分类模型：CNN + 规则融合
3. 多模型融合：提升稳定性
"""
import numpy as np
from typing import Tuple, List
from dataclasses import dataclass
@dataclass
class OccupantInfo:
"""乘员信息"""
classification: str
posture: str
seating_position: str
orientation: str
confidence: float
def classify_occupant_with_camera(
image: np.ndarray,
model_outputs: dict
) -> OccupantInfo:
"""
基于摄像头图像进行乘员分类
Args:
image: 座舱图像, shape=(H, W, 3)
model_outputs: AI模型输出 {
'body_keypoints': 身体关键点,
'height_estimate': 身高估计,
'body_shape': 体型特征
}
Returns:
occupant_info: 乘员信息
Aptiv AOC 方法：
1. 身体关键点检测（身高/体型）
2. 规则层判定（FMVSS 208 法规）
3. 多模型融合（稳定性）
"""
keypoints = model_outputs['body_keypoints']
height_estimate = model_outputs['height_estimate']
if keypoints is None:
classification = "empty"
posture = "na"
elif height_estimate < 1.2:
classification = "child"
posture = detect_posture(keypoints)
elif height_estimate < 0.7:
classification = "infant"
posture = "na"
else:
classification = "adult"
posture = detect_posture(keypoints)
seating_position = detect_seating_position(keypoints)
orientation = detect_orientation(keypoints)
return OccupantInfo(
classification=classification,
posture=posture,
seating_position=seating_position,
orientation=orientation,
confidence=0.95
)
def detect_posture(keypoints: np.ndarray) -> str:
"""
检测坐姿
Args:
keypoints: 身体关键点, shape=(17, 3) [x, y, confidence]
Returns:
posture: normal/forward/side
Aptiv方法：
正常坐姿：肩部关键点水平，脊柱垂直
前倾：肩部前移，脊柱弯曲
侧倾：肩部倾斜，脊柱侧弯
"""
left_shoulder = keypoints[5]
right_shoulder = keypoints[6]
shoulder_tilt = abs(left_shoulder[1] - right_shoulder[1])
hip = keypoints[11]
spine_angle = calculate_spine_angle(hip, left_shoulder, right_shoulder)
if spine_angle < 70:
return "forward"
if shoulder_tilt > 30:
return "side"
return "normal"
def detect_seating_position(keypoints: np.ndarray) -> str:
"""
检测座椅位置（前后位置）
Args:
keypoints: 身体关键点
Returns:
seating_position: front/middle/back
Aptiv方法：
根据膝关键点位置判断座椅前后位置
"""
left_knee = keypoints[13]
right_knee = keypoints[14]
knee_y = (left_knee[1] + right_knee[1]) / 2
if knee_y < 400:
return "front"
elif knee_y < 600:
return "middle"
else:
return "back"
def detect_orientation(keypoints: np.ndarray) -> str:
"""
检测朝向
Args:
keypoints: 身体关键点
Returns:
orientation: forward/sideways/rearward
Aptiv方法：
根据肩部和臀部关键点朝向判断
"""
left_shoulder = keypoints[5]
right_shoulder = keypoints[6]
left_hip = keypoints[11]
right_hip = keypoints[12]
shoulder_width = abs(right_shoulder[0] - left_shoulder[0])
if shoulder_width > 100:
return "forward"
elif shoulder_width > 50:
return "sideways"
else:
return "rearward"
def calculate_spine_angle(hip: np.ndarray, left_shoulder: np.ndarray, right_shoulder: np.ndarray) -> float:
"""
计算脊柱角度
Returns:
angle: 脊柱与垂直方向夹角（度）
"""
spine_mid = (left_shoulder + right_shoulder) / 2
spine_vector = spine_mid - hip
vertical = np.array([0, -1])
angle = np.arccos(np.dot(spine_vector[:2], vertical) / (np.linalg.norm(spine_vector[:2]) + 1e-6))
return np.degrees(angle)
|