Design And Implementation Of Mobile Augmented Reality System Based On Monocular ORB-SLAM With Matching Optimization

In recent years,the short video industry has developed rapidly,and users’ demand for interesting and personalized short video works is also growing day by day.In order to further accelerate the output of high-quality short video works,the application form of short video+emerging technologies has gradually gained attention,among which the combination of short video and augmented reality technology is developing in full swing.In order to give full play to the value of AR virtual-real combination and interactive ability,it is necessary to produce AR interactive short video works based on a realistic environment with richer content,among which the underlying key technology is ORB-SLAM technology based on feature point method.In order to facilitate users to directly create AR interactive short videos using mobile terminals,ORB-SLAM needs to be transplanted to mobile terminals to provide positioning and tracking capabilities.However,for ORB-SLAM with feature point method,feature matching is the key to affect the overall computing time and pose tracking effect.Therefore,in order to better adapt ORB-SLAM to mobile terminals,research focuses on feature matching.In this paper,the feature matching part of the key link of ORB-SLAM is analyzed,and a targeted feature matching optimization scheme is designed.In order to solve the problems of low efficiency and few correct matches of the original matching method in the module of monocular initialization and tracking of reference key frames,GMS feature matching filter is used to efficiently eliminate false matches,reduce the influence of outliers and improve the success rate of operation.In order to solve the problems of low sampling efficiency and long time of multiple iterations of the original RANSAC algorithm in the relocation process of ORBSLAM,a semi-random sampling RANSAC algorithm framework based on quality evaluation function is proposed in this paper,which effectively reduces the number of iterations in the process of model solving,reduces computational overhead and improves the speed of successful relocation.On the whole,the above two matching optimization finally improves the overall trajectory tracking accuracy,ensures the accuracy of position and pose tracking,and maintains the real-time performance and robustness.In this paper,the monocular ORB-SLAM after feature matching optimization is transplanted to the Android mobile terminal,the underlying pose tracking capability is built,and an augmented reality system-AR camera is designed and implemented,which realizes the basic camera function and the augmented reality function based on the reality scene,and can display the interactive effect of AR rendering based on the reality scene in real time.This AR camera helps to improve the efficiency of short video creation in the field of augmented reality.The results of functional test and performance test of the system prove that all functions of the system run normally and the system performance meets the requirements.