基于径向矢量的不变性形状分类方法  

RADIUS VECTORS BASED INVARIANT SHAPE CLASSIFICATION

作　　者：叶芗芸[1] 胡栋梁[1] 戚飞虎[1] 

机构地区：[1]上海交通大学计算机科学与工程系

出　　处：《红外与毫米波学报》1996年第4期250-256,共7页Journal of Infrared and Millimeter Waves

基　　金：国家自然科学基金

摘　　要：介绍一种利用径向矢量提取形状特征的方法，着重于分析二维图形的不变性，提出一种带有方向因子的径向矢量描述．该矢量以图形边界弧长为自变量，完整地刻画了图形特点，克服了非凸图形识别中存在的多义性．以归一化的特征矢量为输入，采用三层神经网络为分类器。A feature extracting method based on radius vectors was introduced.By using a directional factor,the proposed feature gives a complete description of the shapes,and it eliminates the ambiguity that may occur in the traditional shape representation of concave shapes.The normalized features can be used for final classification.With the help of a three layer BP neural network,satisfactory results in recognizing characters and military objects were carried out.

关 键 词：形状分类 径向矢量 不变性识别 神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]