An improved YOLOX detection method for tracking obstacles of unmanned electric locomotives in coal mines under low lighting
To solve the problems such as derailment,collision,and rollover of unmanned electric locomotives caused by insufficient lighting in underground coal mines,a YOLOX-CBAM target detection method considering multiple features of low light is proposed.This method can effectively identify and classify obstacles on the track.First,through the collection and labeling of actual scenes,a dataset of obstacles in coal mines is constructed.Second,the dataset is fed into the Zero_DCE model for data processing.Third,the YOLOX object detection network is improved by adding a dual-channel CBAM attention module in the backbone network and the feature pyramid network respectively,to solve the problem of incomplete data feature extraction by a single channel.Fourth,the loss function in the YOLO header is replaced by SIoU to speed up model iteration.Finally,the CBAM,SA,SA+SIoU,SE,and SE+SIoU modules are added to the YOLOX model to verify the influence of the CBAM attention mechanism and the SIoU loss function on the performance of the YOLOX model.The results show that compared with the two-stage Faster-RCNN network,YOLOv4 network,YOLOv5 network,and the original YOLOX network,the accuracy of the model in this study increases by 4.65 percentage point,2.65 percentage point,2.19 percentage point,and 1.35 percentage point respectively,the recall rate increases by 9.39 percentage point,4.36 percentage point,0.82 percentage point and 0.76 percentage point respectively,and the speed is improved by 28.6 FPS,16 FPS,13.6 FPS,and 2.9 FPS respectively.Compared with the YOLOX model adding CBAM,SA,SA+SIoU,SE and SE+SIoU modules respectively,the accuracy of the model in this study is increased by 0.64 percentage point,0.84 percentage point,1 percentage point,1.29 percentage point,and 0.76 percentage point respectively,and the speed is improved 0.5 FPS,0.4 FPS,0.3 FPS,0.2 FPS,and 0.4 FPS respectively.The method proposed in this study can quickly and accurately detect obstacles on the track of unmanned electric vehicles in coal mines,and provide theoretical support for the intelligent upgrade and safe operation of underground transportation equipment,to reduce the accident rate of coal mining enterprises.
safety engineeringunderground unmanned electric locomotivetarget detectionCBAM attention mechanismSIoU loss function
NETL
NSTL
万方数据
维普
