摘要
针对特征点较少时网格运动统计(GMS)算法存在误匹配集中、鲁棒性差等问题,提出一种基于特征匹配质量评价的GMS特征匹配改进算法。该算法不依赖于增加特征点数量,而是通过引入特征匹配质量因子,依据高质量匹配对计算单应矩阵,剔除误匹配提高匹配准确性,因此具有更好的实用性。实验表明,该算法不仅有效地解决了GMS算法中特征点较少时误匹配集中的问题,而且对特征点数量依赖性更小,算法运行时间相比ORB+RANSAC缩短约30%,算法具有更高的鲁棒性、准确性和实时性。
Aiming at the problems of the GMS algorithm of concentrated mismatching and poor robustness under few feature points,an improved algorithm based on the quality evaluation of feature matching is proposed.The algorithm does not depend on increasing the number of feature points.Instead,it calculates the homography matrix based on the high-quality matching pairs by introducing the quality factor of feature matching,and eliminates the mismatching to improve the matching accuracy.The experiment shows that: 1) The algorithm not only solves the problem of mismatching concentration when there are few feature points in GMS algorithm,but also has less dependence on the number of feature points;2) The running time of the algorithm is shortened by 30% compared with that of ORB + RANSAC;and 3) The algorithm has higher robustness,higher accuracy and better real-time performance.
作者
刘帅
芮挺
王东
杨成松
LIU Shuai;RUI Ting;WANG Dong;YANG Chengsong(PLA Army Engineering University School of Graduate,Nanjing 210000,China;PLA Army Engineering University School of Field Engineering,Nanjing 210000,China)
出处
《电光与控制》
CSCD
北大核心
2021年第7期31-34,共4页
Electronics Optics & Control
基金
国家重点研发计划项目(2016YFC0802904)。
关键词
特征匹配
误匹配剔除
质量因子
单应矩阵
GMS算法
feature matching
mismatching rejection
quality factor
homography matrix
GMS algorithm
