Research on Improved Feature Extraction and Matching Based on ORB
ZHU Jun
ZHOU Jie
Abstract：Visual odometry(VO),as the front end of Visual Simultaneous Localization and Mapping(VSLAM),estimates camera motion based on information between adjacent image frames and then serves the back end.Feature point extraction and matching is the main link of VO.ORB,as a classical feature point extraction algorithm,it provides good real-time performance.However,the cluster distribution of its feature points will lead to poor subsequent matching accuracy.At the same time,the re-al-time performance of its feature point extraction needs to be improved.In this paper,an improved ORB feature extraction and matching algorithm is proposed.Firstly,the key point direction calculation algorithm is improved on the basis of the original ORB to improve the real-time performance.Secondly,the average distribution structure of feature points is introduced to enhance its distri-bution.Finally,the improved feature point matching threshold algorithm combined with RANSAC also effectively eliminates mis-matched points.The experimental results show that the improved algorithm enhances the uniformity of feature point distribution and the real-time of the extraction process,and improves the final matching accuracy.
Keywords：ORBaverage distributionfeature matchingRANSAC
Publication Date：2025-09-20
Online Publishing Date：2025-12-23(First online date of this platform, not the publication date of the document)
Pages：8( 2409-2415,2427 )