Visual SLAM provides navigation and positioning functions for mobile robots. The accuracy of positioning has a profound impact on the quality of the robot's work. It is of great significance to study SLAM positioning. ORB-SLAM2 is a visual SLAM framework with excellent positioning performance. It uses FAST algorithm as feature detection module. After a large number of experiments, this paper finds that the FAST algorithm has the defect of detecting false corners, and believes that ORB-SLAM2 has room to improve positioning accuracy. After analyzing the causes of false corners, this paper proposes an improved FAST corner detection algorithm, called FAST-Ring corner detection algorithm. The FAST-Ring corner detection algorithm constructs a new feature extraction template, and the point is considered as a corner when it meets the template. In order to verify the effectiveness of the FAST-Ring algorithm, a comparison experiment of feature extraction from a single image and a comparison experiment of ORB-SLAM2 positioning accuracy are performed in this paper. It is experimentally verified that the FAST-Ring algorithm proposed in this paper can filter out some false corner points extracted by the original FAST algorithm, and the average absolute positioning error of ORB-SLAM2 using KITTI dataset is reduced by 5 cm compared with the original ORB-SLAM2. The experimental results prove that the performance of the FAST-Ring algorithm is better than the FAST algorithm.
Application of an Improved Fast Corner Detection Algorithm in ORB-SLAM2
Lect. Notes Electrical Eng.
International Conference on Autonomous Unmanned Systems ; 2021 ; Changsha, China September 24, 2021 - September 26, 2021
Proceedings of 2021 International Conference on Autonomous Unmanned Systems (ICAUS 2021) ; Chapter : 180 ; 1825-1833
2022-03-18
9 pages
Article/Chapter (Book)
Electronic Resource
English
Autonomous Navigation Using Monocular ORB SLAM2
Springer Verlag | 2019
|Research on omnidirectional ORB-SLAM2 for mobile robots
IEEE | 2018
|British Library Online Contents | 1998
|