摘要
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像分析与识别的物理依据。传统方法中有基于单个像元波段亮度的、空间纹理的、变换空间的多光谱图像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20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,多光谱图像分析处理便可以在这种特征提取的基础上,研究新的方法。提出一种基于多门限二进制编码的光谱形状描述方法,这种方法的核心就是将多光谱的亮度范围细分成若干个灰度区间,也称为多门限,分别将各波段光谱亮度与多门限进行比较,从而建立一列能够较详细描述光谱形状的二进制编码,我们将这种新的特征描述方法应用于多光谱图像的分类、信息提取和变化检测。遥感图像处理实验的结果表明,这种方法是有效的。
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 single pixel, texture, transformation, however, the essential feature-reflect spectral curve is not described directly in these algorithms. Since 1980'S, a new algorithm of binary encoding was put forward and applied in multispectral match recognition, new algorithm of TM image processing will be found based on this algorithm of feature distill. In this paper, a new algorithm for spectral shape description based on multi-threshold binary encoding is put forward, The key of this algorithm is that the range of terrain mutispectral is devided into several little gray range, what is called 'multi-threshold',each band is compared with several thresholds, then a binary code based on multispectral shape description is set up. And then this new algorithm is apllied in classifiction, information distill and change detect. The new algorithm is proved effective Through remote image experimentation.
出处
《遥感技术与应用》
CSCD
2005年第2期272-277,共6页
Remote Sensing Technology and Application
关键词
地物反射光谱
特征提取
TM图像
多门限
分类
信息提取
变化检测
Reflect spctral of terrain, Feature distill, TM image, Classification, Multi-hreshold, Information distill, Change detect