摘要
特征点的描述在三维物体识别中具有非常重要的意义,针对一般描述子维数过大导致特征点匹配所需的时间、空间消耗过大等问题,提出一种协方差描述子,通过计算特征点邻域的几何特征的协方差矩阵来描述特征点,并将该描述子应用到特征点匹配中.实验结果表明,此描述子不仅能大大减少匹配时间,同时对刚性变换、噪声、采样密度的变化也具有鲁棒性.最后,本文还利用典型相关分析对误匹配点对进行了剔除,获得了良好的特征点匹配效果.
The description of feature points is of great significance in the recognition of 3D objects. For general descriptor, the dimension is so large to cause extreme consumption of time and space on feature point matching, so a covariance descriptor by calculating the covariance matrix of the geometric feature of feature points with their neighborhood points in this paper, and the covariance descriptor is used in the matching of feature points. Experimental results show that this descriptor not only greatly reduces the matching time,but also is robust for rigid transformation, noise and variation of sampling density. Finally,the wrong matching pairwise points are eliminated by the canonical correlation analysis and a good matching effect of feature points is obtained.
出处
《小型微型计算机系统》
CSCD
北大核心
2017年第3期640-644,共5页
Journal of Chinese Computer Systems
基金
山西省自然科学基金项目(2014011018-3)资助
关键词
特征点
几何特征
协方差描述子
特征匹配
典型相关分析
feature point
geometric features
covariance descriptor
feature matching
canonical correlation analysis
