Grain Depot Wearable Device Detection Model Based on YOLOv5

Standardized wearing of helmets and ventilators is the focus of grain depot safety management, which can effectively prevent safety accidents such as falling objects hurting people and fumigation poisoning. This paper is based on the deep learning algorithm to detect the wearing specifications of grain depot operations. Firstly, collect images by simulating the operation scene of the grain depot, recording video frames on the spot, etc. Use Labellmag software to mark the personnel wearing and not wearing the device, and create the original data set; Then, a grain depot wear standard monitoring model based on YOLOv5 is constructed. Add CBAM Attention to improve the network’s attention to local and global features and the ability to extract small target location information. In the post-processing stage, DIOU-NMS is optimized as the screening scheme of the target boxes to reduce the rate of missed detection rate of adjacent targets. Through the contrast experiment, the results show that the improved model mAP(mean average precision) is increased from 92.6% to 93.9%, and finally the detection model meeting the detection accuracy and real-time requirements is obtained.