计算机工程2012，Vol.38Issue(17)：196-200,5.DOI:10.3969/j.issn.1000-3428.2012.17.054
基于Ncut的SIFT特征匹配算法
SIFT Feature Matching Algorithm Based on Ncut
摘要
Abstract
Traditional Scale Invariant Feature Transform(SIFT) feature matching algorithm is based on global search, so bad real-time performance is always shown and mismatches always exist in regions with similar textures. This paper proposes a SIFT feature matching algorithm based on Normalized Cut(Ncut). It puts forward a novel matching method for SIFT feature based on Ncut in the scene of moving objects. It breaks every SIFT feature into its most prominent moving groups and does feature matching respectively between related groups. The proposed algorithm reduces the range of searching and the experimental results show that it provides higher matching efficiency and higher matching accuracy in regions with similar textures. It achieves stable real-time SIFT feature matching between adjacent image frames.关键词
归一化分割/尺度不变特征变换/图论/特征点/特征聚类/特征匹配Key words
Normalized Cut(Ncut)/ Scale Invariant Feature Transform(SIFT)/ graph theory/ feature point/ feature clustering/ feature matching分类
信息技术与安全科学引用本文复制引用
陈抒瑢,李勃,董蓉,陈启美..基于Ncut的SIFT特征匹配算法[J].计算机工程,2012,38(17):196-200,5.基金项目
国家自然科学基金资助项目(61105015) （61105015）
江苏省科技厅科技支撑计划基金资助项目(BE2009667) （BE2009667）
江苏省环境监测科研基金资助项目(0917) （0917）
