Research And Application Of Monocular Mobile Augmented Reality Based On Improved ORB Algorithm

Mobile Augmented Reality(MAR)adds virtual content to the real world through computer technology and presents it through mobile devices.Mobile augmented reality brings people a brand-new interactive way and experience.ORB algorithm is used for feature extraction and matching,and is often used in 3D tracking registration of mobile augmented reality.Because the monocular camera equipped with smart devices can’t calculate the absolute scale,the effect of simultaneous localization and mapping based on monocular vision is poor in some scenes lacking scene texture and fast motion.In order to solve some of the current problems,this thesis studies and tests the monocular mobile augmented reality technology based on improved ORB algorithm,and applies it to the mobile augmented reality system based on monocular vision SLAM technology.Aiming at the higher mismatching rate of ORB algorithm,this thesis proposes an improved ORB feature processing and matching algorithm,which reduces the mismatching rate and improves the matching accuracy.This algorithm uses FAST corner detection algorithm to extract feature points,uses the feature descriptor scheme based on CGRIEF operator to describe the extracted feature points,uses the consistency verification algorithm based on GMS to verify the feature matching results,and uses the improved RANSAC algorithm to eliminate the mismatch of feature points.In this thesis,the effectiveness of the algorithm is verified in TUM public data sets.Compared with the original ORB algorithm,the accuracy of the algorithm successfully matches feature points is improved by 13%.Aiming at the problem that the monocular camera can not calculate the absolute size,a system initialization algorithm based on vision and IMU fusion is proposed.The algorithm aligns visual data with IMU data in a loosely coupled way to recover the initialization system parameters of the camera.In this thesis,the cube evaluation of 3D objects is carried out to build a semantic map,which realizes a more complex occlusion relationship between virtual and real.The robustness of the algorithm is verified on KITTI,a computer vision public data set.The performance of the system using this algorithm is tested,and the system runs stably and normally,giving users a more perfect augmented reality experience.