基于无人机视觉检测的铁路行人侵限实时监测和预警系统

Real-Time Monitoring and Early Warning System for Railway Pedestrian Encroachment Based on UAV Visual Detection

摘要: 本文设计了一种基于无人机视觉检测的铁路行人侵限实时监测和预警系统，结合Yolov5s目标检测算法进行图像预处理及行人侵限行为判断。针对Yolov5s算法，通过引入ECA注意力机制和BiFPN模块，构建Yolov5s-ECB模型，与Yolov5s相比，map提升了2.2%，F1-score提升了10%，检测速度为63张/s，能够很好地满足实际应用的训练和检测要求。因此，本文提出的基于无人机视觉检测的铁路行人侵限实时监测和预警系统具有很好的应用前景和推广价值。

Abstract: This paper designs a real-time monitoring and early warning system for railway pedestrian limit violations based on UAV visual detection, and combines the Yolov5s target detection algorithm with image preprocessing and pedestrian limit violation behavior judgment. For the Yolov5s algorithm, the Yolov5s-ECB model was built by introducing the ECA attention mechanism and BiFPN module. Compared with Yolov5s, the map increased by 2.2%, the F1-score increased by 10%, and the detec-tion speed was 63 pictures/s, which can be very good to meet the training and testing requirements of practical applications. Therefore, the real-time monitoring and early warning system for railway pedestrian intrusion based on UAV visual detection proposed in this article has good application prospects and promotion value.