Safety Helmet and Mask Detection in substation Based on Deep Learning

Safety helmet is the basic protective measure for electric power workers in the construction work. In the current epidemic situation, mask is also an essential protective equipment. But in the complex construction environment, the detection of safety helmet and mask is easy to produce occlusion, resulting in information loss. In order to reduce the safety hazards of power workers and the risk of epidemic transmission, this paper carried out real-time detection for wearing safety hats and masks in power scenarios, the dataset was annotated and enhanced. At the same time, we redesigned the network structure of YOLO V5, the loss function is changed, and the attention mechanism is added at the end of the backbone network. The experimental results show that the proposed method can accurately detect occluded masks and safety helmets with strong generalization in dynamic video detection. The improved algorithm outperforms the original algorithm in terms of accuracy, real-time performance and the size of the network model, and can realize the real-time detection of safety helmets and masks in power construction sites.