针对轻量化网络的安全帽检测方法

变电站内电气设备数量众多,在工人进行现场作业时需要对工人佩戴安全帽进行监测。由于机器学习的安全帽佩戴检测方法常常出现漏检和误检的情况,为提高对安全帽佩戴识别的准确率,同时加快识别速度,提出了一种基于YOLOv5s的轻量化卷积神经网络模型。通过引入RepVGG模块对网络主干进行轻量化,在网络后处理阶段通过Soft-NMS降低遮挡目标漏检率,以Mixup数据增强来扩充数据集,建立样本之间的线性关系,提升训练模型泛化性能,最后进行消融实验。实验结果表明,改进的模型的均值平均精度(mAP)达到80.4%,推理速度达到了83.3 f/s,为变电站安全帽佩戴检测提供了有效参考。

Lightweight Neural Network for Safety Helmet Detection Method

There are a large number of electrical equipment in the substation,so it is necessary to monitor workers wearing safety helmets when they are working on site.Because the helmet wearing detection method of machine learning often has missed detection and false detection,in order to improve the accuracy of helmet wearing recognition and speed up the recognition speed,a lightweight convolution neural network model based on YOLOv5s is proposed,wherein RepVGG module is introduced to lighten the network backbone,Soft-NMS is used to reduce the missed detection rate of occluded targets in the post-processing stage of the network,Mixup data enhancement is used to expand the data set and establish linear relationship between samples,and generalization performance of training model is improved.Finally,ablation experiments are carried out.The experimental results show that the mean average precision (mAP) of the improved model reaches 80.4%,and the inference speed reaches 83.3 f/s,which provides an effective reference for the helmet wearing detection in substations.