基于改进SURF的快速图像匹配算法  

Fast Image Matching Algorithm Based on Improved SURF

作　　者：黄云彬 焦良葆[2] 曹雪虹[1,2] HUANG Yun-bin;JIAO Liang-bao;CAO Xue-hong(School of Communications and Information Technology,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Institute of Artificial Intelligence Industrial Technology,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京工程学院人工智能产业技术研究院,江苏南京211167

出　　处：《软件导刊》2020年第9期220-223,共4页Software Guide

摘　　要：SURF是在SIFT基础上提出的一种图像特征点提取算法。针对传统算法误匹配点多和计算量大等问题,提出一种基于改进SURF的快速图像匹配算法。该算法通过引入对角降维与角度删减方法,分别对SURF算法中特征点描述子进行降维和误匹配点剔除,以提升匹配速度和精确度。实验结果表明,与传统算法相比,该算法提高了1%~10%的匹配正确率,以及8%~30%的效率。Speeded up robust features(SURF)is an image feature point extraction algorithm based on scale-invariant feature transform(SIFT).In order to solve the problems of multiple error matching points and heavy computation in traditional algorithms,this paper proposes a fast image matching algorithm based on the improved SURF.By introducing diagonal dimension reduction and angle reduction methods,the algorithm reduces the dimension of feature point descriptors in SURF and eliminates mismatched points respectively.The algorithm improves matching speed and accuracy.The experiment shows that compared with the traditional algorithm,the proposed algorithm improves the matching accuracy from 1%to 10%and the efficiency from 8%to 30%.

关 键 词：SURF算法 特征点匹配 对角降维 角度删减 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]