Research on identifying illegal behavior of miners riding overhead passenger devices based on M3CFC-YOLOv7-tiny

Abstract

A lightweight miner illegal behavior intelligent detection algorithm based on M3CFC-YOLOv7-tiny is proposed to address the three challenges of deploying lightweight equipment in coal mines, low recognition precision in complex environment, and imbalanced data samples in the automatic recognition task of miners riding overhead passenger devices. MobileNetV3-Small network is introduced into the YOLOv7-tiny model for edge deployment. CBAM attention mechanism is added to enhance the perception and expression ability of miners’ behavior characteristics. Improve the loss function to Focal-CIOU to balance the contribution of sample loss. Conduct ablation experiments and comparative experiments on the self-built dataset. The results show that compared with the original model, the parameters of the improved model are reduced by 30.6%, the floating point calculation is reduced by 46.9%, the detection accuracy is increased by 2.3%, and the model lightweight and real-time detection are realized; compared with 9 object detection models, the improved model has the best comprehensive performance in multiple indicators, and there are no missed detections or false detections.

Keywords

• coal mine safety
• overhead passenger device
• illegal behavior recognition
• m3cfc-yolov7-tiny
• mobilenetv3-small
• cbam attention mechanism
• focal-ciou loss function