Research on Purifying Paired Feature Point Based on RANSAC Algorithm

Abstract

Aiming at the problem that the matching precision of feature points in ORB algorithm is not high enough, an improved feature point matching purification algorithm was proposed. After detecting the feature points of the images, the BF algorithm was used to calculate the Hamming distance between the feature points. Then filtering some wrong matching points through two-way matching, the preliminary matching point pairs were obtained, and then the appropriate distance threshold was set to further filter the matching point pairs, finally using the random sample consensus (RANSAC) algorithm to purify. The experimental results show that the improved feature point matching purification algorithm can improve the matching accuracy under the condition of ensuring the real-time performance of the algorithm.