基于类半径不确定性度量的遥感影像分类

摘要

遥感影像分类是遥感技术应用中的一个重要研究方向，而类准则是分类算法设计过程中的关键问题之一。现有高光谱分类方法很少考虑光谱的不确定性，而地物光谱在获取过程中，由于环境的复杂性，“同物异谱”“同谱异物”现象时有发生，为此该文旨在将光谱特征不确定性分析用于高光谱影像的分类算法中。其主要思想是首先利用类半径进行定量描述光谱的不确定性；然后，通过特征加权的方式将光谱不确定性引入到特征相似性的度量中，并提出一种基于类半径不确定性度量的影像分类方法；最后通过高光谱影像进行分类实验，验证了本文所提算法的有效性。%The classification of remote sensing image is an important topic in the applications of remote sensing technology. The classification rule is one of the significant roles in the procedure of image classification.Noticing that the existed classification method involved the uncertainty of spectrum,and in the process of obtaining a spectral,the phenomenon of “same thing different spectrum”and “foreign matter same spectrum”appear frequently.Therefore,a high-spectral image classification method based on the measure of spectrum is described in this paper.The proposed method can be divided into the following steps:firstly,the uncertainty of spectrum is measured by using the radius of categories;and then,a classification method is constructed by introducing the spectrum uncertainty into the measure of similarity;at last,the effectiveness of the proposed method is verified by experiments.