汽车工程学报2026，Vol.16Issue(1)：104-112,9.DOI:10.3969/j.issn.2095-1469.2026.01.09
基于视觉与建模的乘员姿态识别及损伤研究
Driver Posture Recognition and Injury Analysis Based on Vision and Modeling
摘要
Abstract
Road traffic safety is a critical global public safety issue,and driver postures have a significant impact on crash injuries.Therefore this study proposes a driver posture recognition and reconstruction method that integrates visual perception and simulation modeling.Specifically,in-vehicle cameras are used to capture occupant images,after which the OpenPose model extracts posture information and converts it into joint angles.These angles then drive the Madymo multi-body model to conduct frontal crash simulations and obtain injury data.The results show that the recognition confidence for three posture types(normal driving,forward-leaning,and reclined)exceeds 0.8,and significant differences are observed in the 15-ms head injury criterion(HIC15)and chest compression values.This study provides theoretical and technical support for the future integration of driver monitoring and active safety technologies.关键词
乘员姿态识别/姿态分析/OpenPose/MADYMO/智能预警/主动安全Key words
occupant posture recognition/posture analysis/OpenPose/MADYMO/intelligent warning/active safety分类
交通工程引用本文复制引用
常佳龙,吴磊,匡高远,李泉,尚诗,张帆,王青春,唐亮..基于视觉与建模的乘员姿态识别及损伤研究[J].汽车工程学报,2026,16(1):104-112,9.基金项目
国家自然科学基金项目(52402441) （52402441）
外国专家项目(H20240157) （H20240157）
中央高校基本科研业务费专项资金资助项目(BLX202336) （BLX202336）
