一种优化的特征点匹配算法

An optimized feature point matching algorithm

为提高三维重建的性能,提出一种优化特征点匹配算法。首先,通过组合算法对特征点进行初匹配;其次,通过随机抽样一致算法剔除误匹配获得目标的精确匹配点,并将精确匹配点通过区域生长算法,经多次迭代实现稠密匹配;最后,经过二次剔除误匹配点得到精确稠密匹配结果。同时,搜索窗的大小通过局部二值模式来自适应调整,提高区域生长后新增种子点的精度,从而更有利于三维重建。实验证明该方法能提高匹配效率和精度。

An optiinized feature point matching algorithm is proposed to improve the performance of 3D reconstruction. Firstly, the feature points are matched by combination algorithm. Secondly, the accurate matching points are selected after eliminating mismatches by RANSAC method and then processed by adaptive dense matching based on region growing. Finally, the accurate matching points are obtained after twice eliminating mismatches. Meanwhile, the local binary pattern is adapted to search window, which can increase the precision of new seed points after region growing and therefore lay a good foundation for 3D reconstruction. The experimental results show that the matchingefficiency andaccuracy can be improved.