摘要
针对传统的SURF局部特征匹配算法实时性不高的问题,充分利用ORB特征点检测算法简单高效的优势,提出了一种新的特征点匹配算法。首先,针对原始ORB特征匹配算法出现大量误匹配对的问题,采用基于K最近邻的特征点描述后,对前后两帧特征点进行双向匹配,再通过顺序抽样一致性算法进一步提纯。实验结果表明,经过本文算法提纯后匹配对准确度提升到99.9%,平均耗时0.46 s,处理速度约是SURF特征匹配算法的5倍,SIFT特征匹配算法的25倍,能够满足实时运用的需求。
As traditional SURF local feature points matching algorithm has low real-time,a new feature points matching algorithm is proposed by using the advantage of ORB feature points detection. First of all,there are a large number of false matching pairs in the original ORB,and these false matching pairs are described by feature points based on K nearest neighbor,and then the feature points in two consecutive frames are bilaterally matched. At last,the consistency is further refined by sequential sampling algorithm. Experimental results show that the match accuracy is up to99. 9% after purification,and the average running time is 0. 46 s. Processing speed is about 5 times faster than SURF feature matching algorithm and 25 times faster than SIFT feature matching algorithm,and it can meet the requirements of real-time processing.
出处
《激光与红外》
CAS
CSCD
北大核心
2015年第11期1380-1384,共5页
Laser & Infrared
基金
国家自然科学基金项目(No.61301233)
全军军事学研究生课题(No.213JY512)资助