一种基于多门限二进制编码的多光谱图像分类新方法

A New Classification Algorithm for Multispectral Image Based on Multi-threshold Binary Encoding

不同类型的地物具有不同的反射光谱 ,在多维光谱空间中构成不同的特征向量 ,这便是遥感多光谱图象分析与识别的物理依据。传统方法中有基于单个象元波段亮度的、空间纹理的、变换空间的多光谱图象特征提取与分析方法 ,但这些方法并没有直接描述地物的最本质特征———反射光谱曲线 ,从上个世纪 80年代开始 ,当二进制编码的方法提出并在多光谱匹配识别中获得成功应用后 ,多光谱图象分析处理便可以在这种特征提取的基础上 ,研究新的方法。本文提出一种基于光谱形状描述的多门限的二进制编码分类方法 ,并给出了图象实验的结果。

Different types of terrain would lie on the different site in the spectral space, which is the basic theory for remote analysis and recognition. Commonly, feature distill and analysis of mutispectral image is based on sigle pixel,texture,transformation, however, the essential feature reflect spectral curve is not described directly in these algorithms. Since a new algorithm for binary encoding for description of spectral shape is put forward since 1980's, a new algorithm of multispectral image processing based on Description of Spectral Shape will be found. In this paper, a classification algorithm based on multi-threshold binary encoding for Description of Spectral Shape is introduced, and result of image experimentation is present.