Research On Target-tracking Algorithm Based On The Ar Auxiliary Maintenance System Of EMU

With the rapid growth of the passenger volume and mileage of the Chinese high-speed railway,the daily maintenance workload of Electric Multiple Unit(EMU)is increasing year by year.It is an important task to protect safe operation of EMU and use the EMU maintenance information to complete the maintenance work rapidly,accurately and timely.In this dissertation,tracking algorithm for Augmented Reality(AR)based on Tracking-Learning-Detection(TLD)was proposed on the basis of research for AR technology.And the AR auxiliary maintenance system of CRH380BL was designed to assist maintenance personnel to complete maintenance tasks,integrate maintenance information and regularize operating process,so as to improve the efficiency of daily maintenance work.In this dissertation,system target detection algorithm,target tracking algorithm and 3 Dimensional(3D)registration algorithm combining the characteristics of CRH380BL EMU maintenance operation scene after research of basic principle for target tracking algorithm based on computer vision were designed.This research focus on achieving high speed in the calculation process under the premise of ensuring the high accuracy of target tracking results.The work completed in this dissertation are as follows:Firstly,feature extraction algorithm was studied,and Histogram of Oriented Gradient(HOG)was proposed as the feature extraction algorithm.Based on the feature extraction experiment of the operation scene,feature extraction of CRH380BL EMU target tracking was completed,and the feature matching experiment was carried out.After that the advantages and disadvantages of the target detection algorithm were analyzed.Secondly,TLD tracking algorithm based on target detection was improved by hybrid programming of MATLAB and VS2013.And the effectiveness of algorithm was verified by taking the video data of CRH380BL EMU bogie's replacement and maintenance as an example.Combined with Kernel Correlation Filter(KCF)algorithm,the TLD-G target tracking algorithm was designed for AR system.The original TLD algorithm pays more attention to the detection results of detector.However,the sliding window detector needs to process a large number of image blocks,which seriously affects the real-time performance.The method of reducing the scanning range of sliding window was proposed to improve the computational speed of the algorithm.The standard data set experiment of TLD-G algorithm and the video data experiment of replacement operation were carried out.The results show that the speed,accuracy and robustness of the algorithm can meet the requirements of AR system.Finally,the AR auxiliary maintenance system of CRH380BL EMU was designed and tested.The design of AR auxiliary maintenance system of CRH380BL EMU was completed in this part,and the system maintenance work field test was carried out.The test results show that the system has a good performance in real-time and 3D registration accuracy.