基于极坐标区间运算的2D形状匹配  

2D Shape Matching Based on Polar Interval Operations

作　　者：张国敏[1] 殷建平[1] 祝恩[1] 毛玲[2] 

机构地区：[1]国防科学技术大学计算机学院,长沙410073 [2]国防科学技术大学电子科学与工程学院,长沙410073

出　　处：《计算机研究与发展》2008年第z1期286-290,共5页Journal of Computer Research and Development

基　　金：国家自然科学基金项目(60373023)

摘　　要：形状匹配是遥感图像目标识别、字符识别、手形识别和步态识别等任务中的关键步骤之一.针对刚体识别任务中形状匹配易受方向、尺度和位置等仿射变化量影响的情况,提出了一种新的基于极坐标区间运算的2D形状匹配算法.该算法首先以形状区域的中心点为极点,区域的最长轴方向为极轴,对形状区域进行归一化的极坐标变换;然后定义了同一角度对应的区域内点区间之间的运算;最后定义了两个区域归一化极坐标变换结果在区间运算下的相似度函数,用以表征两个区域之间的匹配度.从可见光遥感图像中提取的实物图像实验结果证明,该方法能够有效归类相似形状,并能区分各类不同的形状.Shape matching is one crucial step in the recognition of remote sensing image objects, characters, hand shapes, and gait, etc. In order to eliminate the effects of rigid object affine variables such as orientation, scale, and position, a novel shape matching algorithm is proposed based on polar interval operations. First, normalized polar coordinate translation is carried out in the shape region with an origin at the middle of the region and a polar axis of the longest region orientation axis. Then, interval operations are defined between inner point intervals of the region at the same orientation. Finally, a similarity function on the normalized polar translations of two shape regions is defined based on the interval operations. It indicates the similarity of the two shapes. Experiments on actual objects in visible light remote sensing images demonstrate that the algorithm in this paper can sort out similar shapes and distinguish different shapes effectively.

关 键 词：形状匹配 极坐标变换 区间运算 相似度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]