Research On Wellbore Inspection System Of Deep Vertical Shaft Based On Machine Vision

As an important passage for mines,deep shafts not only need to transport mining equipment and minerals,but also need to transport workers,so the safety of deep shafts is particularly important.However,with the increase of mining depth and the change of crustal stress,defects such as deformation of tank channels,wall peeling or cracks of the well wall may be caused,affecting the normal operation of tank cages and other equipment,and mine safety accidents may occur.In order to be able to monitor various problems in the wellbore in time,collect and analyze the situation in the wellbore in real time,and prevent the occurrence of dangerous accidents,this paper designs a set of intelligent monitoring system,which uses the tank inspection robot to carry out patrol inspection to achieve effective detection of the wellbore status.Machine vision technology is used to detect tank defects and provide support for subsequent wellbore health evaluation and maintenance.The main research contents of this paper are as follows:(1)A high-safety inspection robot was designed,and an inspection scheme was designed with this inspection robot.Firstly,based on the actual working environment of the wellbore in the coal mine,a new structure of tank inspection robot is designed,which uses STM32F103 as the main control chip,installed with power loss protection device,higher safety,can run back and forth along the surface of the tank road,and collect various index data in the wellbore in a circular manner.(2)An improved YOLOv5s network was designed.The existing YOLOv5s network cannot run smoothly in the Raspberry Pi due to its large number of parameters and calculations.The improved YOLOv5s network has significantly reduced the amount of parameters and calculations,and the performance indicators are basically unchanged.(3)The improved and YOLOv5s network was deployed to the Raspberry Pi 4B.In order to run the improved YOLOv5s network more smoothly,the second generation neural computing stick of Intel is used in the Raspberry Pi to accelerate the detection speed,and after acceleration,the detection speed is significantly improved to meet the detection needs.(4)Make data sets and design host computers.The image label production tool is used to label the canist defect image,and the canister defect dataset is made and used to train an improved YOLOv5s network that meets the requirements of this paper.At the same time,the upper computer is designed to display the data collected by the tank inspection robot and organize the data statistics,and when the tank road defect is detected,it can also be displayed in the middle of the visual interface and given a red letter warning.Figure [71] Table [16] Reference [82]...