In view of the shortcomings of SLAM indoor scene mapping based on RGB-D vision, such as vulnerable to sunlight, narrow field of vision, large noise, etc., the accuracy and robustness are poor in practical application. Meanwhile, the traditional RGB-D method does not consider the filtering of redundant of view field. Based on Kinect and 2D lidar, an indoor mapping and positioning method with low cost, high accuracy and strong reliability is proposed in this paper. Firstly, 3D point cloud is obtained by PCL processing, and the rotation and translation vectors of camera are obtained by ORB feature point matching method to realize point cloud splicing. Secondly, point cloud down-sampling is carried out, and Generate 2D grid map. The actual trafficability of robot is taken into account to filter out the influence of obstacles in the range of passing height. Finally, the Bayesian reasoning is adopted to achieve the map construction under multi-sensor information fusion. The simulation results show that the algorithm proposed is better than the traditional ORB-SLAM2 algorithm in real time application, which improves the calculation efficiency significantly. It also outperforms to reduce the possibility of collisions with obstacles during the AVG driving process.
Research and Application of Improved VSLAM Algorithm
Lect. Notes Electrical Eng.
2021-10-30
10 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch