Abstract: In contemporary urban living, ensuring the safe operation of public transportation has become an important issue. Recently, a series of bus accidents in China have sparked widespread societal concern, with studies indicating that driver abnormal behavior is one of the main causes of such accidents. In response to this challenge, this paper integrates technologies such as deep learning and video analysis to specifically analyze driver abnormal behavior in bus scenario based on selfdeveloped intelligent monitoring platform. The Intel? Movidius TM MyriadTM 2 module is integrated into the bus prototype to build a core computing unit for video processing and driver abnormal behavior analysis. With the help of high-performance image recognition algorithms, this system can monitor and assess driver operational status in real time. By detecting and correcting inappropriate behaviors, this system can be used for daily training in public transportation companies and provide research data for driver driving situation awareness research platforms.

Key words: fatigue driving detection; experiment teaching platform; bus; deep learning; virtual simulation

摘要： 在当代城市生活中，确保公共交通安全运营已成为一项重要课题。近期，我国发生的 多起公交事故引发了社会广泛关注，研究表明，驾驶员异常行为是导致此类事故的主要原因 之一。为应对这一挑战，文章基于自主开发的智能监控平台，融合深度学习、视频分析等技 术，专门针对公交车场景下的驾驶员异常行为展开分析。将 Intel? Movidius TM MyriadTM 2 模 块集成至公交车样车，搭建起一套视频处理和驾驶员异常行为分析的核心计算单元，借助高 性能的图像识别算法，该系统能够实时监测并评估驾驶员的操作状态，一旦发现异常情况即 可触发预警机制。通过及时发现和纠正驾驶员的不当行为，可以用于公交公司日常培训，为 驾驶员驾驶态势感知研究平台提供研究数据。

关键词: 疲劳驾驶检测；实验教学平台；公交车；深度学习；虚拟仿真