Shape matching algorithm based on shape contexts
Shape matching algorithm based on shape contexts
复制标题
基于形状上下文的形状匹配算法
DOI：
10.1049/iet-cvi.2014.0159
复制
发表时间：
2015
影响因子：
1.7
通讯作者：
Huang Baoqi
中科院分区：
文献类型：
--
作者：
Zhao Long;Peng Qiangqiang;Huang Baoqi
提出了一种利用形状上下文的形状匹配算法。该算法的主要贡献在于：（1）提出了一种新的基于形状上下文的形状匹配算法框架，与现有方法不同的是，提出了一种基于多项式拟合的特征点提取方法作为预处理步骤，以提高基于形状上下文的形状描述子的性能;（2）设计了一种基于卡方统计测度的投票分类方法，用于评价匹配结果。实验结果表明，该方法能够实现高性能，即使测试对象的形状遭受平移，旋转和缩放。
This study proposes a novel shape matching algorithm through exploiting shape contexts. The contributions of the proposed algorithm are twofold: (i) a new framework is presented to deal with the shape matching problem based on shape contexts, but differently from existing methods, the authors exploit a polynomial fitting-based feature point extraction method as a preprocessing step, so as to enhance the performance of the shape contexts-based descriptor; (ii) the authors design a voting classification method based on the chi-square statistical measure to evaluate the matching results. The experimental results show that this method is able to achieve high performance, even if shapes of testing objects suffer from translation, rotation and scaling.