Abstract: With the excellent adaptability, flexibility and efficiency, trackless tyred vehicle has gradually become the core component of modern coal mine auxiliary transportation. Due to the poor lighting conditions and complex scene along the auxiliary transportation line in coal mine, the accidents of the trackless tyred vehicle occur frequently, seriously threatening the life safety of operators along the auxiliary transportation line. Aiming at the problem of reliable identification of intruders in front of underground trackless tyred running along the auxiliary transportation line of coal mine, a reliable identification method of personnel along the auxiliary transportation line based on infrared/visible light dual-mode fusion technology was proposed. In the improved method, an infrared/visible dual-mode fusion network based on multi-scale feature interaction module is constructed, which combines multi-scale fusion and cross-scale fusion to effectively preserve the structural details of infrared images and visible images by designing the window-based multi-head self-attention and shifed window-based multi-head self-atention. The simulation results show that in processing infrared/visible light fusion, the proposed dual-mode fusion technology has higher peak signal-to-noise ratio, structural feature similarity index, mean square error and correlation index than two commonly used image fusion networks such as the Dif-Fusion and SwinFusion. The post-processing method of YOLOv8 target recognition network is optimized, the DIoU-NMS structure is used to improve the accuracy of target positioning, reduce the missed detection rate of the algorithm, and realize the reliable identification of auxiliary transportation tunnel operators. With the improvements, the reliable identification of auxiliary transportation roadway operators is realized. Based on the actual images of the coal mine scene and the processing expansion, an effective data set for the detection task of personnel along the auxiliary transportation in the coal mine is established. The industrial experiment were carried out in Wangjialing Mine of China Coal Huajin Group Co., LTD., and the results show that the accuracy rate of personnel detection and the recall rate of intruders can reach 97.58% and 98.53% respectively.