A Novel YOLOv5-based Anomalous Object Detection Algorithm in Buses
Shida Liu, Qingyi Li, Honghai Ji, Li Wang, Xiaoping Zhang, Zhonghe He · 2022 IEEE 11th Data Driven Control and Learning Systems Conference (DDCLS) · 2022
In this work, a novel abnormal objects analyzing and detecting (AOAD) algorithm inside the bus is proposed, and the AOAD algorithm is further applied to the practical bus by designing an embedded video analysis system. The proposed algorithm is based on the deep learning YOLOv5 (You Only Look Once) algorithm, which has better timeliness and accuracy in detecting anomalous objects. The anomalous objects with larger boxes are detected by building an anomalous object dataset for training. The effectiveness and applicability of the proposed algorithm is verified by extensive experiments on video data based on real buses.