基于有序点集距离的形状匹配算法  

Shape Matching Algorithm Based on Distance of Ordered Points

作　　者：朱鹏伟 李勃[1] 王云烨[1] 陈启美[1] 

机构地区：[1]南京大学电子科学与工程学院,南京210023

出　　处：《上海交通大学学报》2015年第12期1826-1832,共7页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金项目(61105015);江苏省自然科学基金项目(BK2010366);江苏省科技厅项目(BE2011747)资助

摘　　要：针对基于豪斯多夫距离的匹配算法存在时间复杂度高、较难处理部分匹配和无法精确对位等问题,提出了一种改进的基于有序点集距离的形状匹配算法.该算法利用轮廓点集的有序性,动态计算点集之间的匹配关系.算法首先计算曲线的轮廓点曲率,并根据匹配代价作出匹配矩阵,然后基于匹配矩阵的连通情况来判断2幅图像是否匹配,最后依据最短连通路径附近的像素点分布来确定最终的匹配点.实验证明,本算法能加快匹配速度,较好地处理部分曲线匹配,并能确定匹配轮廓点到点的一一对应关系.A shape matching algorithm was proposed based on the distance of ordered points, which used the orderliness of the contour points to calculate the matching relationship between the contours dynamically. First, the algorithm calculated the curvature of every point and drew the matching matrix according to the matching cost. Then, the fact that whether the two images match or not was determined based on the connectivity of the matching matrix. In the end, the final matching points were obtained by analysing the pixel distribution near the shortest communication path. Realistic experiments prove that the method can accelerate the matching speed, handle partial matching well, and match every point of each contours accurately by.

关 键 词：形状匹配 曲率 有序点集距离 匹配矩阵 豪斯多夫距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]