改进的ORB特征匹配算法

Improved ORB feature matching algorithm

针对传统ORB算法对边缘敏感且误匹配率较高的不足,提出了一种改进的ORB特征匹配算法。首先,采用改进的FAST14-24算法进行角点的初步提取,克服了传统ORB角点检测算法对图像边缘敏感的不足;然后在角点精筛选阶段,使用Shi-Tomasi角点检测算法剔除大部分的伪角点,得到重复率更高的特征点集;然后,使用改进的类视网膜描述符提取算法提取512维二进制描述符,并对描述符进行训练,使描述符具有更好的区分性;最后,在特征匹配阶段使用汉明距离进行匹配。在图像匹配实验中,本文算法的匹配正确率相比传统的ORB算法提高了18%~62%。实验结果表明:所提算法具有比ORB算法更高的匹配精度以及更好的鲁棒性。

Aiming at the shortcomings of traditional ORB algorithm which is sensitive to edges and has high mismatch rate,an improved ORB feature matching algorithm is proposed.Firstly,the improved FAST14-24 algorithm is used to extract the corner points,which overcomes the shortcomings of sonsitivity of the traditional ORB corner point detection algorithm to image edges.Then,in the corner point fine selection stage,the Shi-Tomasi corner point detection algorithm is used to eliminate most of the pseudo corner points,and the feature point set with higher repetition rate is obtained.Then use the improved retinal descriptor extraction algorithm to extract 512-dimensional binary descriptors,and train the descriptors to make the descriptors more distinguishable;Finally,the Hamming distance is used for matching in the feature matching phase.In the image matching experiment,the matching accuracy of the algorithm is improved by 18%~62%compared with the traditional ORB algorithm.The experimental results show that the proposed algorithm has higher matching precision and better robustness than the ORB algorithm.