ORB-Oriented Mismatching Feature Points Elimination
Cai-Wei Shu, Xuezhong Xiao · 2018
In order to solve the problem of low accuracy of feature point matching in Oriented FAST and Rotated BRIEF (ORB) algorithm, a mismatching points elimination algorithm is proposed. For the pairs of descriptor obtained by brute force matching based on ORB, the angle cosine method was used for preliminary screening, and the Random Sample Consensus (RANSAC) algorithm based on homography matrix was used for further mismatching points elimination. Experimental results showed that the proposed algorithm can effectively eliminate mismatching points. On the basis of high efficiency, the matching accuracy was also greatly improved compared with the original ORB algorithm. The proposed algorithm is suitable for applications which require high matching accuracy and real-time performance.