Monocular vio (Visual-Inertial Odometry) based mobile terminal AR method

Abstract

The invention provides a monocular vio (Visual-Inertial Odometry) based mobile terminal AR method, which comprises the steps of performing image acquisition and inertial data acquisition respectivelyby using a camera and an IMU; extracting feature points of each image acquired by the camera, and setting the extracted feature points to be key frames; calculating corresponding IMU position and rotation information of two images in a pre-integration manner; acquiring a tracked feature point in the key frames, and calculating the pose of each key frame according to the feature point; calculatingan external parameter between the camera and the IMU according to a condition that displacement and rotation information acquired by the IMU is equal to displacement and rotation information measuredby the camera between the two frame; initializing the speed, the acceleration and a scale factor of the IMU; acquiring the pose of the camera by using tight coupling nonlinear optimization, and then superimposing a virtual object to achieve an AR effect. The mobile terminal AR method adopts the IMU to make up defects of VSLAM (Virtual Simultaneous Localization and Mapping), and is not affected byenvironmental feature points within a short time; and deviation is prevented by using alignment of the VSLAM and the IMU when the time is long.

Description

technical field [0001] The present invention relates to an AR method, in particular to a mobile terminal AR method based on monocular vio. Background technique [0002] SLAM (simultaneous localization and mapping) is divided into two functions: positioning and mapping. Among them, the main role of map building is to understand the surrounding environment and establish the corresponding relationship between the surrounding environment and space; positioning is mainly to judge the pose of the device in the map based on the built map, so as to obtain the information of the camera in the environment. [0003] SLAM is divided into dense and sparse. Based on the consideration of computing power, the mobile terminal generally chooses sparseSLAM. Although Sparse SLAM is fast, the effect is very dependent on the environment: in a complex environment with many feature points, sparse SLAM works well, but for environments with fewer feature points such as white walls and smooth planes,...

AI Technical Summary

Problems solved by technology

Although Sparse SLAM is fast, the effect is very dependent on the environment: in a complex environment with many feature points, sparse SLAM works well, but for environments with fewer feature points such as white walls and smooth planes, the effect is very poor. It is difficult to accurately estimate the pose of the camera

Images