Research on the detection algorithm of electric workers in the limited spaces of hydroelectric power station

Abstract

Safety incidents in limited spaces during power production frequently occur, often leading to disrupted schedules and, in severe cases, substantial economic losses. Traditional detection methods for electric workers in the limited spaces of hydroelectric power stations suffer from limitations such as insufficient feature expression capability and low accuracy. To address these issues, this paper proposes a novel detection method based on an improved You Only Live Once (YOLO) model combined with person re-identification. Firstly, the YOLO model is optimized by enhancing the Backbone network, Neck network, and Head network, and by incorporating a screening and matching algorithm based on anchor-free boxes. Subsequently, by integrating person re-identification, an improved YOLO model fused with the C-JDE model (FD-YOLO) is proposed to enhance the detection accuracy of electric workers in the limited space. Experimental results show that the average accuracy indicators, AP50 and AP50:95, of the proposed model have both increased by more than 12 percentage points compared to the traditional YOLO model, which demonstrates that the improved FD-YOLO model significantly enhances the detection accuracy of electric workers in the limited spaces of hydroelectric power stations. The research outcomes provide valuable technical support for the intelligent management of power stations.