Power Engineering Safety Helmet Detection Algorithm Based on BR-YOLOv7

In the construction of power engineering, wearing a safety helmet is of vital importance. However, due to the problems of dense multi-scale and mutual occlusion of the detection targets during the safety helmet detection process, there are situations such as false detection and missed detection in the detection process. In response to this situation, this paper proposes the BR-YOLOv7 object detection algorithm. First, to enhance the network's feature extraction ability for safety helmets, this paper introduces the BiFormer attention module. This module combines the feature extraction mechanisms of Transformer and BiFPN, enabling the network to better capture the feature information of safety helmets, improving the network's ability to extract the features of safety helmets, thereby enhancing the accuracy and stability of the detection model in complex environments. For the mutually occluded safety helmet detection targets, this paper introduces the repulsion loss function and optimizes it to reduce the loss during the bounding box regression process. Experimental results show that the safety helmet detection accuracy of this algorithm reaches 97.7%, which is 1.9% higher than the original algorithm and meets the requirements of power engineering for the detection accuracy of safety helmets.