A new ORB feature matching image algorithm based on Non-maximum suppression method and Retinal sampling model
Yongfu Liu, Youming Wang · 2021 International Conference on Control, Automation and Information Sciences (ICCAIS) · 2021
Traditional oriented FAST and rotated BRIEF (ORB) feature matching algorithm has mismatch problems in complex environment. To solve the problem, a new ORB feature matching algorithm is proposed. First, Non-maximum suppression method is used to eliminate the feature points and the retained feature points are described by retinal sampling model. Then, the feature points are matched by Hamming distance. Finally, a progressive sampling consensus (PROSAC) algorithm performs matching pair purification and calculates the matrix transformation. Experimental results show that the matching accuracy of the proposed ORB algorithm in complex illumination environment is greatly improved compared with the traditional ORB algorithm. It has been demonstrated that the proposed ORB algorithm possesses strong robustness and real-time performance.