Indoor localization technology of SLAM based on binocular vision and IMU

Abstract

In order to enhance the localization accuracy of mobile robot, a SLAM system combining binocular vision and IMU sensor is proposed in this paper. The IMU sensor is added on the basis of ORBSLAM2 framework. During system initialization, binocular vision and IMU are initialized separately, and then the visual frame is converted to the gravity coordinate system through the rotation matrix between the world coordinate system and the gravity coordinate system, so as to eliminate the influence of gravity on IMU. Finally, the nonlinear optimization was used to optimize the visual constraint and IMU constraint at the same time to complete the fusion of binocular vision and IMU. The simulation experiment was carried out in RuRoc. The results show that the positioning accuracy of the proposed scheme is superior to the ORBSLAM2 system in different experimental scenarios.