Research on Miner Smoking Recognition Method Based on YOLOv7
Smoking behavior of underground miners seriously affects the production safety of coal mines,and effective recognition of underground miners'smoking behavior is imminent.Aiming at the special environment of underground coal mines and the problem of low accuracy of traditional recognition methods,it proposes a YOLOv7-based miners'smoking behavior recognition algorithm YOLO-SFN.SimAM is embedded into the network structure of YOLOv7,and the Focus module is used to replace the 3×3 convolution kernel in the lower branch of MPConv,so as to improve the model's feature extraction ability in the complex background.Soft-NMS is used as the post-processing algorithm for the network model in the post-processing stage,which solves the leakage detection problem of the traditional NMS algorithm in the complex and dense environment.The experimental results show that the accuracy rate of the method is 96.45%,the recall rate is 92%,and the precision rate is 97.05%.The research results have been popularized and applied in Chensilou coal mine,realizing the effective supervision of the smoking behavior of miners in underground coal mines.
国家科技期刊平台
NSTL
万方数据
