Application Research Of Mobile Augmented Reality Technology Based On ORB-SLAM

Augmented reality technology is dedicated to superimposing computer-generated objects onto real-world scenes.It has become an important field in computer vision research,and augmented reality on mobile devices is gradually appearing in the user's field of vision.The use of visual SLAM positioning technology to achieve augmented reality has always been the focus and difficulty of SLAM research.The SLAM of binocular cameras and RGB-D cameras imposes severe hardware requirements on mobile devices.However,monocular cameras are ubiquitous in modern mobile devices,so the research of monocular vision SLAM system positioning technology on mobile devices has a good application market.The research content of this paper mainly focuses on the positioning technology of monocular vision SLAM,and uses it to realize the augmented reality on the Android platform.This paper first introduces the basic principle of positioning of monocular vision SLAM system,obtains camera internal parameters and eliminates distortion by calibration,through feature point extraction and matching,2D-2D motion solution,3D-2D motion solution,3D-3D motion to solve camera motion.And space point location.Next,according to the engineering source code,the specific steps of the monocular positioning technology of the modern visual SLAM system-ORB-SLAM are analyzed.This article uses the second iterative version of ORB-SLAM,which also becomes ORB-SLAM2.It mainly analyzes two important aspects of the positioning process: camera motion tracking and relocation of tracking failure.After analyzing the positioning principle and the specific implementation of ORBSLAM2 positioning,this paper continues to do the following work:(1)The ORB feature point algorithm is improved,which is aimed at the problem of incorrect extraction of feature points caused by over-aggregation of ORB feature points and excessive local brightness.In the ORB feature point extraction process,local adaptive threshold acquisition and local optimization are added.Feature point acquisition steps,and experiments were compared to their effects(2)Porting ORB-SLAM2 on the Android platform and optimizing its initialization steps.Since ORB-SLAM2 is not specifically for mobile devices,this article ported it to Android devices,and optimized the initialization process of monocular ORB-SLAM2 on Android devices in order to improve the user experience of Augmented Reality on Android phones.Reduce the initialization threshold,reduce the calculation method that is not suitable for mobile devices,improve the initialization success rate,and finally conduct experiments to compare optimization effects.(3)Apply the research content to the substation inspection training to build an Android augmented reality application.This is because the power system is large in scale,numerous in variety,complicated in division of labor,fast in updating,and strict in requirements for practitioners.The use of augmented reality technology to display maintenance equipment in three-dimensional space can bring real experience to employees,and can be used for training at any time and place.