Year: 2025
Type: article
Abstract: The behaviour of electric bicycle into the lift is a great security risk, but also one of the causes of many fires in the building, bringing a serious threat to people's life and property safety. In this paper, we use YOLOv8 model to identify and detect e-bikes in the lift scene. Compared to YOLOv5, YOLOv8 improves performance and employs an anchor-free mechanism, reducing hyperparameter tuning complexity and enhancing small object detection in elevator scenarios. The experimental test results show that: the mAP@50 reaches 87%, and the operating speed reaches 125 FPS, which meets the deployment requirements of the actual monitoring system. This study provides a new solution in the field of lift safety monitoring, which can significantly improve the efficiency of detecting e-bikes in lifts and prevent safety accidents such as fires caused by e-bikes.
DOI: 10.54097/9e6t0280
Authors: Xuanjie Xu, Shuxiang Yang, Haoran Shu
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4408829607