Research On Intelligent Detection Technology Of Unmanned Trafficability Of Monorail Crane Based On Machine Vision

In recent years,China has begun to accelerate the process of intelligent mining construction.As one of the important pieces of transportation equipment for underground auxiliary transportation in coal mines,monorail cranes have the advantages of small size,large carrying capacity,and strong climbing ability.Studying the unmanned driving technology of monorail cranes is an inevitable direction for current and future intelligent development.However,the research on unmanned driving technology for monorail cranes is still in its early stages,especially given the lack of intelligent detection technology research for the underground environment of coal mines.This thesis focuses on the special environment of underground tunnels in coal mines.Starting from improving the safe passage of monorail crane unmanned driving,research is conducted on the detection technology of monorail crane unmanned driving targets and the detection technology of monorail crane track joint distance.This has important research significance for promoting the development of unmanned monorail crane driving technology.The main research contents of this thesis are as follows:Firstly,analyze the feasibility and difficulties of unmanned target detection and track joint detection for monorail cranes,and select the sensors required for intelligent detection of unmanned trafficability of monorail cranes.Secondly,target detection algorithms based on candidate regions and regression are compared and analyzed.Based on the YOLOv3 algorithm,a YOLOv3 detection algorithm with an improved channel attention mechanism and loss function is proposed,and the improved YOLOv3 algorithm is tested and analyzed.The results show that the accuracy of the improved YOLOv3 algorithm is improved by 9.1%,the m AP is improved by 3.6%,and the detection speed can reach 35.6 FPS.Once again,a machine vision based joint detection algorithm for monorail crane tracks was proposed,which mainly includes image preprocessing,histogram information statistics,morphological processing,etc.The distance between track joints was calculated based on calibration coefficients.The results showed that the machine vision based joint detection algorithm for monorail crane tracks took approximately 36 ms to process a joint image,with a detection error of 0.3 mm.Then,a monorail crane was built to simulate the tunnel environment,and unmanned target detection experiments were conducted on the monorail crane.The detection effects of Faster R-CNN,YOLOv3,and the improved YOLOv3 algorithm proposed in this thesis were compared.A comprehensive evaluation was conducted in terms of the AP value,P-R curve,F1 curve,etc.The results showed that the average accuracy of the improved YOLOv3 algorithm reached 90.4%,with higher detection accuracy and better overall performance.Finally,a human-machine interaction system for detecting the seam of a monorail crane track was designed,and a large number of monorail crane track seam images were collected for track seam detection experiments.The actual distance and detection distance of 10 sets of seam samples were compared.The results showed that the proposed monorail crane track seam detection algorithm had an average processing time of 34.6 ms,an average detection error of 0.73 mm,and a maximum detection error of no more than 1.1 mm,indicating high accuracy,and the advantage of fast detection speed.In summary,the monorail crane unmanned target detection algorithm and monorail crane track joint detection algorithm designed in this thesis can improve the safe passage of monorail crane unmanned driving and can also promote the unmanned and intelligent coal mine development process to a certain extent.There are 77 figures,13 tables,and 115 references in this thesis.