期刊文献+

一种基于SIFT的改进优化特征匹配算法 被引量:5

An optimized image feature matching algorithm based on SIFT
下载PDF
导出
摘要 尺度不变特征变换即SIFT算法存在实时性差,易误匹配等固有问题,本文针对性地提出了特征描述符降维处理和匹配优化解决方案,得到一种能满足更高实时性和精确性需求的特征匹配算法。通过使用特征点为中心的9个同心圆环梯度累计值,构建72维特征向量,进行特征描述符降维,达到简化特征描述的目的,从而减少描述符的生成和匹配时间。此外,结合匹配点择优筛选和RANSAC算法匹配提纯,有效地减少了误匹配。实验表明:改进优化后的特征匹配算法既显著地提高了特征匹配精确度,又改善了算法自身实时性。 The scale invariant feature transform(SIFT)algorithm shows some deficiencies of nature,such as poor real-time performance and easy mismatching.In view of this,a solving scheme of dimensionality reduction in the feature descriptor as well as matching optimization for image features is proposed so that a feature matching algorithm which can meet the higher requirements of real-time and matching accuracy is obtained.Among it,the 72 dimensional feature vector is constructed by using the gradient cumulative values of 9 concentric rings centered on the feature point to reduce the dimension of the feature descriptor,so as to simplify the feature description and reduce the time of descriptor generation and matching.Furthermore,by combining the optimal selection of matching points and the matching purification of RANSAC algorithm,the mismatching is effectively reduced.Experiment shows that the optimized algorithm not only significantly improves matching accuracy,but also betters real-time in itself.
作者 甘小红 覃志东 蔡勇 肖芳雄 GAN Xiaohong;QIN Zhidong;CAI Yong;XIAO Fangxiong(School of Computer Science and Technology,Donghua University,Shanghai 201600,China;Shanghai UAZAN Precision Technology Co.Ltd.,Shanghai 201600,China;Software Engineering School,Jinling Institute of Technology,Nanjing 211169,China)
出处 《智能计算机与应用》 2021年第11期5-9,共5页 Intelligent Computer and Applications
基金 国家自然科学基金(6126200) 2020国家级大学生创新创业项目(112-03-0178010/001)
关键词 SIFT算法 特征描述符 匹配点择优筛选 RANSAC算法 SIFT algorithm feature descriptor matching point selection RANSAC
  • 相关文献

参考文献3

二级参考文献16

共引文献11

同被引文献42

引证文献5

二级引证文献11

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部