Design And Implementation Of SLAM Based On Mobile Phone Multi-Camera In Augment Reality APP

The concept of "metacomes" is emerging,and augmented reality(AR)technology is described as a physical portal to the metacomes,a technology that combines virtual information with the real world.The real-time location and scene awareness function is the premise,SLAM technology is one of the important implementation algorithms of environment awareness.At the same time,due to the popularity of mobile phones,the improvement of mobile computing performance and the support of multisensor,mobile terminal SLAM technology has shown a high application value.However,due to the limitation of computing power of mobile terminal,there are many problems in directly transferring SLAM technology to mobile terminal.This paper designs and implements an augmented reality system suitable for mobile phones.The main work of this paper mainly includes:In this paper,an algorithm VOrbSLAM is proposed.Before the optical flow calculation is introduced into the key frame selection,the selection method based solely on the feature point results is optimized through the optical flow result selection,so as to reduce the computation amount and speed up the front-end processing speed of SLAM.Compared with OrbSlam,the speed can be improved by 62.5%,and the accuracy is between the feature method and the direct method.The average absolute pose error tested on the four sub-data sets selected by Euroc data set is 0.055.In this way,the defects of feature-point SLAM calculation and the inaccuracy of optical flow method are solved,and an acceptable balance between accuracy and speed is found for mobile terminals.Based on the research on the influence of camera field of view on SLAM algorithm,and by taking full advantage of the features of multiple cameras in mobile phones,this paper proposes a sliding window method,which uses the main camera and wide-angle camera to enhance the robustness of SLAM algorithm and reduce the probability of tracking loss without increasing too much computation and minimizing the precision of map construction.On the homemade data set,Movement speed increased by more than 25%when tracking is lost.This paper designs and implements a set of augmented reality system that can run on Android terminal,and applies it in the field of intelligent home decoration.The VOrbSLAM algorithm proposed in this paper is transplanted to Android devices,and combined with the augmented reality rendering engine,the multi-threaded framework is used to optimize the algorithm so as to reach the real-time standard.At the same time,user interaction is added to assist the algorithm to better build the map.More refined perceptual environment to improve user experience.