燃气站背景下基于YOLOv8改进的防护装备检测研究

Gas station as a key node of the gas system, the degree of its staff protection standardization has an important impact on the safety of the station. In order to reduce the risk of accidents at the gas station, it is necessary for the staff to wear helmets in accordance with the norms. In this paper, we improve the YOLOv8 model based on the YOLOv8 model, and for the problems of low pixel and small proportion of helmet in the gas station monitoring video, we propose to replace the backbone network with SPD-Conv to improve the target detection ability for low quality images and small targets, and at the same time, we experimented with different loss functions, and finally obtained the improved target detection model with loss function of SIoU and backbone network of SPD-Conv, and confirmed the effectiveness of the above improvements through experiments. detection model, and experimentally confirmed that the above improvement effectively improves the target detection effect for helmets.