Identification of miners' unsafe behaviors based on transfer learning and residual network

Abstract

In order to accurately identify unsafe behaviors of miners and reduce occurrence of accidents in coal minesꎬ an image recognition method combining transfer learning and deep residual network is proposed. Firstlyꎬ behavior instances of miners were divided into three dimensionsꎬ namely completely safe behaviorsꎬ relatively safe behaviorsꎬ and unsafe behaviorsꎬ among which completely safe behaviors included walkingꎬ sitting and standingꎬ relatively safe behaviors included bendingꎬ squattingꎬ liftingꎬ pushingꎬ pullingꎬ waving and clappingꎬ and unsafe behaviors included falling and throwing. Thenꎬ ResNet50 network was used for trainingꎬ and transfer learning weight parameters of ImageNet data set were fine ̄tuned. Finallyꎬ 12 classification was conducted through full connection layerꎬ and final classification results were checked against test data. The results show that residual network model based on transfer learning is superior to other deep neural network models in identification accuracy of falling and throwing movementsꎬ and it can effectively identify unsafe behaviorsꎬ thus avoiding accidents caused by human factors.