Intelligent Helmet Detection System based on the Improved YOLOv5

In the train maintenance workshop, the standardized wearing of helmets by maintenance personnel can effectively avoid a larger part of safety accidents, so it is extremely significant to monitor the wearing of helmets by overhaul personnel in real time. In this paper, we propose an improved YOLOv5s helmet detection algorithm based on the deep learning method, and deploy it to the edge-end devices to realize a set of intelligent monitoring system for the helmet wearing of the maintenance personnel. In this paper, two improvement measures are proposed for the YOLOv5s algorithm. Firstly, for the problem of helmet misidentification and missed identification, the convolutional block attention module (CBAM) is incorporated in three different positions of the Backbone part of YOLOv5s to enhance the extraction capability of Backbone for important features and improve the recognition accuracy of the model for helmets; secondly, since workers’ head areas are often obscured by each other in real scenes, replacing GIoU Loss with CIoU Loss in YOLOv5s can effectively improve the convergence speed of the network training and the regression localization accuracy of the network. The comparison experimental results fully prove the effectiveness of the proposed method.