基于K-means聚类改进的特征匹配算法  

Improved Feature Matching Algorithm Based on K-means Clustering

作　　者：苗璐欣 杨德宏[1] MIAO Luxin;YANG Dehong(Faculty of Land and Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China)

机构地区：[1]昆明理工大学国土资源工程学院,云南昆明650093

出　　处：《软件导刊》2024年第11期129-133,共5页Software Guide

摘　　要：图像的匹配和融合是图像拼接的核心步骤,而在图像匹配过程中,传统的KNN匹配算法中错误匹配较多,且图像上的正确匹配点对连线的斜率几乎平行。鉴于此,使用K-means算法对其进行聚类以筛选出正确匹配。实验表明,该算法可以去除大量错误匹配,较传统算法匹配率最大可提升近1倍,匹配时间最大可缩短约0.2 s。在图像融合中对最佳缝合线算法作出改进,并且与渐出渐入融合算法进行比较实验,证明改进的缝合线算法可有效去除图像重叠区域的“鬼影”、错位现象。Image matching and fusion are the core steps of image mosaic.In this paper,in the process of image matching,there are many wrong matches in the traditional KNN algorithm.It is observed that the slope of the correct matching point on the image is almost parallel,and the K-means algorithm is used to cluster and screen out the correct matching.Through experiments,the algorithm can remove a large number of false matches,and the matching rate and matching time are improved.The optimal suture algorithm is improved in image fusion,and com⁃pared with the gradual fusion algorithm,it is proved that the improved suture algorithm can effectively remove the ghost and dislocation phe⁃nomenon in the overlapping area of the image.

关 键 词：KNN匹配 Kmeans聚类 RANSAC算法 最佳缝合线 渐出渐入算法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]