首页 | 本学科首页   官方微博 | 高级检索  
     

基于拟Laplace谱的形状表示与聚类
引用本文:唐俊,梁亮,梁栋,朱明. 基于拟Laplace谱的形状表示与聚类[J]. 医学教育探索, 2011, 0(6): 749-753
作者姓名:唐俊  梁亮  梁栋  朱明
作者单位:安徽大学计算智能与信号处理教育部重点实验室,合肥 230039;安徽大学计算智能与信号处理教育部重点实验室,合肥 230039;安徽大学计算智能与信号处理教育部重点实验室,合肥 230039;安徽大学计算智能与信号处理教育部重点实验室,合肥 230039
基金项目:国家自然科学基金(11071002);安徽省教育厅自然科学研究项目(KJ2011A008);安徽大学211 工程学术创新团队项目
摘    要:基于谱图理论的形状表示与聚类是计算机视觉和模式识别领域的重要研究方向。针对不同形状的结构特征,通过对形状骨架点所构完全图的拟Laplace矩阵进行奇异值分解,将得到的高维数据投影至低维空间中,进而分析该数据在低维空间中的分布情况实现聚类。针对公共数据集的对比实验验证了该算法的有效性。

关 键 词:聚类; 形状表示; 图; 拟Laplace谱
收稿时间:2011-04-12

Shape Representation and Clustering Based on Quasi Laplace Spectrum
TANG Jun,LIANG Liang,LIANG Dong and ZHU Ming. Shape Representation and Clustering Based on Quasi Laplace Spectrum[J]. Researches in Medical Education, 2011, 0(6): 749-753
Authors:TANG Jun  LIANG Liang  LIANG Dong  ZHU Ming
Affiliation:Key Laboratory of Intelligent Computing and Signal Processing, Ministry of Education, Anhui University, Hefei 230039, China;Key Laboratory of Intelligent Computing and Signal Processing, Ministry of Education, Anhui University, Hefei 230039, China;Key Laboratory of Intelligent Computing and Signal Processing, Ministry of Education, Anhui University, Hefei 230039, China;Key Laboratory of Intelligent Computing and Signal Processing, Ministry of Education, Anhui University, Hefei 230039, China
Abstract:Shape representation and clustering based on spectral graph theory is a hot topic in the field of computer vision and pattern recognition. Aiming at the structure features of different shapes, the high dimensional data are obtained by means of singular value decomposition on quasi Laplace matrices of the skeleton of shapes. Furthermore, the shapes are clustered by analyzing the distribution of the projection in a low dimensional space. The comparative experiments on the public data set demonstrate the effectiveness of the proposed approach.
Keywords:clustering   shape representation   graph   quasi Laplace spectrum
点击此处可从《医学教育探索》浏览原始摘要信息
点击此处可从《医学教育探索》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号