A Rear and Side Object Detection System for Autonomous Driving Bus
Xinwei Wang, Yingquan Zou · 2023
To guarantee the safety of autonomous buses when changing lanes, this paper improves and implements a reliable and cost-effective object detection system that suits low-speed automatic driving buses. In this system, the ECX-1400-PEG embedded industrial control computer served as the primary computing unit. The detecting system and control unit communicates via the Redis database. Our System's detection part uses YOLOv5-Rep, an improved network whose kernel size is as large as$31\times 31$. Experimental results show that compared with YOLOv5s, the number of YOLOv5-Rep's parameters was reduced to 6.67M, and the$\text{F}_{1}$score increased to 0.9236. The system can meet the detection requirements of the rears and sides of low-speed autonomous driving buses when changing lanes.