摘要
傅立叶描述子是分析和识别物体形状的重要方法之一.基于雷达图表示多维数据的原理,提出了一种利用傅立叶描述子识别雷达图形的可视化数据分类新方法.该方法采用多元统计中的雷达图表示多维数据,不同模式类别的多维数据构成不同形状的雷达图多边形.在此基础上,给出基于极半径函数的傅立叶描述子来描述和识别雷达图的边界曲线特征.运用概率神经网络,以傅立叶描述子为输入特征向量完成自动识别雷达图形.实验结果表明这种分类方法有良好的分类精度,可与传统分类器性能相比.
Fourier descriptor is an important method used in shape analysis and recognition. A novel method for designing the classifier of multi-dimensional data was proposed, which used radar chart of multi-statistics to show multidimensional data and applied Fourier descriptors to recognize the radar chart. Different multi-dimensional data formed different radar chart and distinguished different category. Then a new Fourier descriptor based on polar radius is defined, which describes curve of radar chart shape. The method of Probabilistic Neural network combined with Fourier descriptors is used to implement automatic classification. Experimental results show this method has the good classification precision, and may compare with the traditional classifier.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2010年第1期178-183,共6页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(60671025
60474065)
关键词
数据可视化
雷达图
傅立叶描述子
形状识别
概率神经网络
data visualization
radar chart
Fourier descriptors
shape recognition
probabilistic neural network
