SLAM BASED ON INFORMATION FUSION OF STEREO VISION AND ELECTRONIC COMPASS

摘要

Due to the effects of image resolution, camera calibration and work environment in stereo vision-based robot's SLAM (Simultaneous Localization and Mapping), there are certain problems in location accuracy, robustness and anti-jamming capability. Meanwhile, because pose estimation is achieved through incremental iteration, there is also cumulative error. In this paper, an integrated location algorithm based on stereo vision and electronic compass is proposed to improve accuracy and robustness through information fusion. The rotation angles obtained by electronic compass and stereo vision respectively are fused by an improved fuzzy adaptive extended Kalman filter. Then initial pose estimation is obtained by the rotation angles and 3D coordinates of the time t and the time t + 1, and accurate pose estimation is realized by an adaptive particle filter. Finally, the map is updated by a Kalman filter. Experiment results show that the location accuracy, robustness and real-time performance are better than stereo vision alone and the traditional fuzzy adaptive extended Kalman filter.