基于谱图理论的形状聚类仿真(英文)

Simulation of Shape Clustering Based on Spectral Graph Theory

谱图理论已经广泛的应用于形状表示、匹配和聚类。主要研究两个方面的问题:1、通过提取形状的骨架特征点和图谱分析,建立形状的数学模型;2、通过分析模型的数学特征,将形状投影到低维空间中。利用上述方法实现了形状的聚类仿真。针对公共数据集的仿真实验证明了该聚类算法能有效的实现形状聚类。

Spectral graph theory has been exploited widely for the purpose of shape representation,matching and clustering.The proposed algorithm focuses on two problems： Firstly,skeletal feature points and spectrum were used to build up the mathematic model of the shape.Secondly,shapes were organized into a low dimensional space by analyzing the mathematic character of models.Meanwhile,the simulation of clustering was achieved by researching the distribution of shapes in this space.Simulation experiments on the public data sets illustrate that the presented algorithm is available for shape clustering.