基于AKAZE的BOLD掩码描述符的匹配算法的研究

MATCHING ALGORITHM OF BOLD MASK DESCRIPTOR BASED ON AKAZE

针对现有基于KAZE/AKAZE的改进算法大多不具有良好的鲁棒性问题,提出基于AKAZE的BOLD掩码描述符的匹配算法。将该算法与尺度不变算法、ORB算法、AKAZE应用于放射变化、尺度变化、压缩比变化、模糊变化的待匹配图像测试集中,对匹配正确率和匹配时间进行对比实验。结果显示,该算法的匹配正确率相较原AKAZE提升了10%左右,匹配速度仅比原AKAZE慢15%左右。该算法在提升AKAZE鲁棒性的同时,匹配速度下降不明显,可应用于对鲁棒性和匹配速度较高的匹配场景。

In order to solve the problem that most of the existing improved algorithms based on KAZE/AKAZE do not have good robustness,this paper proposes a matching algorithm based on AKAZE BOLD mask descriptor.This algorithm was compared with SIFT algorithm,ORB algorithm and AKAZE in radiation change,scale change,compression ratio change,and fuzzy change to match the image test set,and comparison experiments were performed on the matching accuracy rate and matching time.The experimental results show that the matching accuracy of our algorithm is about 10%higher than the original AKAZE.And the matching speed is only about 15%slower than that of the original AKAZE.It is shown that the algorithm improves the robustness of AKAZE,while the matching speed is not significantly reduced.And it can be applied to the matching scene with high robustness and matching speed.