Design And Implementation Of Safety Detection System For Mining Belt Conveyor Based On Deep Learning

Mine belt conveyor is an important equipment in the underground work of coal mine.Its safety detection is mainly to monitor the abnormal conditions of conveyor belt in real time,identify the targets in the image and mark the positions,which is of great significance to ensure the safe production of coal mine industry.However,at present,there are some problems in the detection research of mine conveyor belt,such as missing dataset,single detection category and insuffi cient sample size.In addition,the special environment of dim light and dust floating in underground coal mine makes it difficult to distinguish targets from the background,which brings difficulties to object detection task.In view of the lack of dataset,a special image dataset MCBD(Mine Conveyor Belt Data)is made for mine conveyor belt.To solve the problem of single detection category,image data is collected from different angles,and the range is selected from the belt head and the area around belt conveyor.Taking into account various abnormal conditions of the conveyor belt,three target categories are marked:workers around the conveyor belt,no-load status of the conveyor belt,and large materials on the conveyor belt.Aiming at the problem of insufficient sample size,the RandomMix method is proposed to expand the original dataset in the form of multiple random mixing.Using YOLOv5 and YOLOv7 as the test algorithms,the average detection accuracy of the enhanced dataset is 10.3%and 17.7%higher than that of the original dataset,which proves that the RandomMix data augmentation method has a positive effect on the object detection task.In view of the difficulty in distinguishing targets from the background caused by special environment in underground coal mine,an improved algorithm called YOLOCA is proposed.Based on the single-stage object detection model YOLO,the coordinate attention mechanism is added to the Backbone network structure,making full use of position information to enhance the feature representation of targets,so that the network pays more attention to the target area in the whole image,thus improving the detection accuracy.The experimental results show that the detection accuracy of YOLOCA is improved by 1.3%and 5.3%compared to YOLOv5 and YOLOv7,respectively,which verifies the effectiveness of the improved algorithm.Based on the above research work and achievements,the safety detection system for mining belt conveyor is designed and implemented.This system integrates multiple functions such as model processing,object detection,information display,image defogging,etc.,and is suitable for the underground working environment of coal mine.Applying the improved YOLOCA algorithm to the object detection module of the system not only meets the requirements of real-time monitoring,but also has high detection accuracy.After testing,the detection time of a single image is 20.1ms,and the FPS reaches 49.75,which verifies the feasibility of the system.