行人异常越界行为识别的嵌入式系统设计

Embedded System for Pedestrian Abnormal Crossing Behavior Detection

在现代化的居家、商业、工业等领域中,越界行为已经成为一种常见的安全风险。为解决现阶段越界行为识别算法存在实时性差、难以适应复杂场景等问题,设计行人异常越界行为识别嵌入式系统,通过YOLO算法进行目标检测,利用ByteTrack多目标跟踪算法跟踪关联检测到的目标,使用区域监测算法结合多目标跟踪算法的跟踪轨迹精准识别越界行为。实验结果表明,本系统能够部署在搭载Aidlux系统的智能手机上,在不同场景下实现对人体越界行为的准确识别,并发出警报信号,满足了大部分场景的应用需求。

In modern fields such as home,commerce,and industry,cross-border behavior has become a common security risk.To address the issues of time-consuming performance and difficulty in adapting to complex scenes in the current algorithm for identifying pedestrian abnormal crossing behavior,an embedded system for identifying pedestrian abnormal crossing behavior is proposed.The YOLO algorithm is used for object detection,and the ByteTrack multi-objective tracking algorithm is used to track and associate the detected targets.The area monitoring algorithm is combined with the tracking trajectory of the multi-objective tracking algorithm to achieve accurate identification of crossing behavior.The experiment results have shown that the proposed system can be deployed on smartphones equipped with the Aidlux system,achieving accurate recognition of human behavior beyond boundaries in different scenarios and issuing alarm signals,meeting the application requirements of most scenarios.