The Research And Application Of Key Technologies For Intelligent Operation And Maintenance Of Rail Transit Based On Mixed Reality

The rapid development of the rail transit operation and maintenance has increased the workload of daily operation and maintenance of railway equipment..The existing rail transit operation and maintenance have problems,such as large equipment size,complex internal structure,difficulty in operating and moving,and difficulty for first-time operators to quickly and comprehensively understand the equipment.Mixed reality,as an emerging technology,can utilize its diverse interactive methods to create a seamless immersive experience and realistic visual effects for users,which can effectively assist in the rail transit operation and maintenance.However,due to the complex changes in the real environment of rail transit,there are still some problems in the process of mixed reality operation and maintenance of rail transit,such as target tracking is vulnerable to object occlusion and environmental impact,pose estimation is vulnerable to background clutter and large amount of model calculation,which can easily lead to poor stability of the mixed reality rail transit operation and maintenance system.In response to the above issues,the main research work of this article is as follows:(1)A twin network-based target tracking algorithm for railway complex scenes is proposed to address the problems of target susceptibility to occlusion or deformation,missing attention points in feature extraction and low feature accuracy in twin network-based target tracking algorithms.By replacing the feature extraction network with Res Net,replacing traditional cross correlation with pixel by pixel correlation,and adding a attention mechanism based on standardization,the network can accentuate areas that account for a large proportion of feature weights.At the same time,a template update subnet is improved to address the impact of similar interference on target tracking during the tracking process,effectively improving the accuracy of the algorithm.(2)Aiming at the limited computing power of existing mobile portable devices,as well as some chaotic backgrounds and the lack of correlation between different levels of features,an improved pose estimation algorithm based on Pose CNN is proposed based on the research of target tracking algorithms in complex railway scenes.By replacing the convolutional block in the network feature extraction module with a more weight Ghost module,and adding a pyramid pooling module to the semantic tag branch of Pose CNN,feature fusion is performed on different pooling layer features,which enhances the network’s ability to capture information about objects with complex backgrounds and the correlation between their context information,and effectively improves the ability of object posture estimation.(3)Based on target tracking and pose estimation algorithms,a hybrid reality based intelligent rail transit operation and maintenance system was developed by combining hybrid reality technology with rail transit operation and maintenance.The system mainly includes functions such as automatic adsorption of equipment components,component perspective observation,component disassembly and assembly operations,component location prompts,equipment knowledge learning,and remote video calls,which will effectively improve the efficiency of traditional rail transit operation and maintenance.