基于极线约束的ORB特征匹配算法

图像匹配是机器视觉领域的基础核心课题，针对当前ORB (Oriented FAST and Rotated BRIEF)图像特征匹配算法虽然执行速度快，但是匹配质量不高的问题，提出一种通过极线约束来改进ORB匹配的算法。通过合理设计Hamming阈值大小来提高初始匹配点数量，采用RANSAC和8点改进法计算基本矩阵，应用极线约束剔除误匹配保留大量优质匹配点。仿真实验结果证明，算法改进后的优质匹配点数量可达原始算法的2-3倍，同时极大地提高了匹配点的质量，证明了算法的有效性。

ORB feature matching based on epipolar constraint

Image matching is the core of the field of machine vision. Addressing the problem that the existing ORB feature matching algorithm is fast but of low matching quality, this paper proposed an improved version of the ORB algorithm for improving matching accuracy further by enforcing the epipolar line constraint. By reasonably designing the Hamming threshold to get a large of original matching points and using RANSAC and 8-point improved algorithm to get fundamental matrix, the algorithm improved the number of matching points. Using epipolar line constraint to eliminate the wrong match, it was able to retain a large number of high quality matching points. Experiment results show that the algorithm can increase the number of matching points to 2-3 times of the original algorithm and greatly improve the quality of matching points, which proves the effectiveness of the algorithm.