Target Tracking Algorithm For AR Aided Assembly Process

Locomotive maintenance process is an important part of railway transportation.The structure of locomotive system is complex,the workload of maintenance process is large,and the requirements for technical personnel and working equipment are high.In order to solve the problem of high maintenance frequency and low information level in the process of locomotive bolt assembly,it is necessary to develop a set of locomotive bolt auxiliary assembly system composed of software,intelligent wrench,AR glasses and working terminal.In the process of AR assisted assembly of locomotive bolt assembly system,the maintenance personnel need to wear AR glasses.AR glasses use the mark points generated by the target tracking algorithm to track the assembly bolts.The maintenance personnel use the intelligent wrench to assemble the bolts according to the instructions of the mark points,and record the tightening torque of the intelligent wrench and the corresponding bolt number into the database.Aiming at the problem that the mark points of the target tracking algorithm used in the current system are difficult to accurately track the current bolts under the interference of the factors such as the fast movement of the AR eyeglasses’ view angle,the occlusion of the intelligent wrench,etc.,this paper studies the existing long-time target tracking algorithm LCT and proposes an improved algorithm to achieve the accurate and stable target tracking for the AR assisted assembly process.In this thesis,the target tracking algorithm is divided into three parts: feature extraction,motion model and detection model,and the influence of each part on the performance of the algorithm is explored through the control variable method.The experimental results show that the long-time target tracking algorithm LCT is suitable for the target tracking of locomotive bolt AR auxiliary assembly system.Based on the specific analysis of the occlusion situation of the target,the APCE index is introduced to improve the original occlusion discrimination mechanism and model update strategy on the basis of LCT algorithm,and SVM classifier is used to detect and relocate the target and solve the occlusion problem of the target,the CN feature of color histogram is introduced to realize the adaptive weight distribution of the hog and CN features by using the peak proportion of response graph,and a target tracking algorithm LCT＿RK for AR assisted assembly process is designed to eliminate the AR Based on the influence of the fast movement of the eyeglass camera’s view angle on the tracking results.In this thesis,LCT＿RK algorithm is applied to the locomotive bolt assembly system,and the interference factors such as occlusion and fast moving of visual angle are tested in the actual scene.The results show that the real-time performance of the system is 20 ～ 25 FPS,which meets the real-time requirements of the system.Moreover,the mark points generated by the algorithm can be accurately determined to the bolts to be assembled,which guarantees the assembly quality of the locomotive bolt assembly process.