关键词: 机器视觉, 同时定位与地图构建, 自主移动机器人, 八叉树地图, 避障

Abstract: Abstract: Mobile robot is faced with the problem of simultaneous positioning and 3D reconstruction when exploring unknown environment, and the feature-based VSLAM algorithm can accurately estimate pose and create 3D sparse map, which is not conducive to the application of robot. Based on the octree map technology, this paper improves the map construction method of ORB-SLAM, constructs the 3D octree map suitable for robot application in real time, and solves the problem that sparse map can’t be used for obstacle avoidance and navigation, and the validity of the method proposed in this paper is verified using the authoritative dataset. Running the same datasets, 3D positioning errors and running time are 25% and 56% of the RGBD-SLAM method, respectively. Finally, this paper builds an autonomous mobile robot, and applies the improved VSLAM system to the mobile robot, which can realize the autonomous obstacle avoidance and the 3D map construction in real time, which provides a useful reference for the practical application of VSLAM technology.

Key words: machine vision, simultaneous localization and mapping, autonomous mobile robot, octree map, obstacle avoidance