基于YOLO v5的工人玩手机行为检测方法研究

Research into Detection Method of Workers Mobile Phone Playing Behavior Based on YOLO v5

为规范工人生产行为、减少安全事故发生,提出一种监控工人使用手机行为检测算法.该算法以YOLO v5模型为基础,对其网络结构和损失函数进行改进.首先,优化主干网络,将ConvNeXt Block和SPP结构引入浅层网络增加浅层特征的提取;然后,在主干网络与特征聚合网络之间构建CBAM注意力机制层,过滤冗余信息;最后,选取EIoU损失函数代替GIoU损失函数,提高模型收敛速度与检测结果的定位精度.通过自建工人使用手机行为数据集,分别对YOLO v5原模型、改进模型以及主流模型进行对比.试验结果表明,在人体和手机目标检测中,改进模型有更好的检测精度和检测速度.

In order to regulate the production behavior of workers and reduce safety accidents,a behavior detection algorithm for monitoring workers using mobile phones is proposed.Based on the YOLO v5 model,this algorithm can be used to improve the network structure and loss function.Firstly,the backbone network is optimized by introducing ConvNeXt Block and SPP structure into the shallow network to increase the extraction of shallow features;secondly,a CBAM attention mechanism layer is constructed between the backbone network and the feature aggregation module to filter redundant information;finally,EIoU loss function is selected to replace the GIoU loss function for improving the convergence speed of the model and the positioning accuracy of the detection results.The original YOLO v5 model,the improved model and the mainstream model are compared by means of workers mobile phone behavior data set collected in this study.The experiments show that the improved model is better in detection accuracy and speed in human and mobile target detection.