基于泊松方程-Fourier-Mellin矩的形状识别和分类方法  

Shape Recognition and Classification Based on Poisson Equation-Fourier-Mellin Moment

作　　者：邹建成[1] 客楠楠[1] 鲁燕[1] 

机构地区：[1]北方工业大学图像处理与模式识别研究所,北京100144

出　　处：《北方工业大学学报》2009年第1期1-10,共10页Journal of North China University of Technology

基　　金：国家自然科学基金(No.10671002);北京市自然科学基金(No.4002011);北京市人才强校计划项目基金资助项目

摘　　要：提出一种新的形状描述子,即泊松方程-Fourier-Mellin矩描述子.首先求解定义在形状区域上的泊松方程,再利用泊松方程的解构造特征函数,然后将这些特征整体化得到Fourier-Mellin矩,用泊松方程-Fourier-Mellin矩描述子来表示形状.这种描述子改进了LenaGorelick提出的泊松方程—几何矩描述子,它不但保留了泊松方程—几何矩描述子既能代表形状整体特征,又能描述形状的部分特征的优点,又结合了Fourier-Mellin矩的优势,对形状的描述更加全面和准确.最后的形状识别和分类实验证明,我们提出的泊松方程-Fourier-Mellin矩描述子的效果更好.A new shape descriptor is introduced in the paper, which is also known as Poisson Equation-Fourier-Mellin moment descriptor. First, we solve the Poisson Equation in the shape area, and use the solution to get the feature function, which are then integrated using Fourier-Mellin moment to represent the shape. This method develops the Poisson Equation-geometric moment descriptor proposed by Lena Gorelick, and keeps both the advantages of Poisson Equation-geometric moment and Fourier-Mellin moment. It is proved that it is better than other two descriptors in shape recognition and classification experiments.

关 键 词：泊松方程 Fourier-Mellin矩 形状识别 形状分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]