Application Research of Safety Helmet Detection Based on Low Computing Power Platform Using YOLO v5

Nowadays, long-distance and small target detection has become a research hotspot in computer vision. There are various potential dangers in the construction site with complex environment, and safety accidents often occur because of not wearing safety helmet. Due to the large number of personnel and the changeable environment, the traditional manual inspection and video surveillance have some problems, such as poor detection efficiency and lack of timeliness. In this paper, we propose an improved YOLO v5 method and deployed on Raspberry Pi. Firstly, an attention mechanism is introduced to solve the problem that the backbone network is not sensitive to feature differences; Secondly, the loss function is improved and GFocal Loss is used to train the model. In this paper, the helmet is carried out on the Raspberry Pi, and the experimental results show that the improved YOLO v5 algorithm is better than the original algorithm to detect the target, which is helpful to the practical deployment of intelligent transportation, traffic flow and other application scenarios.