Abstract:At present,the disassembly and assembly of the bolts of the rail freight car hook pallet are carried out manually.There are problems such as high work intensity,high safety hazards,low efficiency,and difficulty in controlling the tightening torque.There is an urgent need for automation and intelligent equipment.Based on the disassembly and assembly scenes and process requirements of the hook pallet bolts,a mobile intelligent disassembly and assembly robot was developed to replace manual for bolts automatic disassembly and tightening.A machine vision positioning system for bolts was designed to complete the automatic positioning of all bolts on the hook pallet.The YOLOv5 network was used as the bolt detection and positioning model,and the network was trained on a self-made〖JP〗 data set.The trained target detection model was used to locate the bolts of the hook plate in real time to realize the servo control of the robot.The bolt positioning test was carried out on the hook plate bolt simulation physical platform.The test results show that by the method,the automatic bolt positioning can be realized efficiently and reliably,and it has strong robustness to bolts with different surface conditions.