多点模拟及其在遥感影像信息提取中的应用及进展

Multiple-point Simulation and Its Application and Development in RS Image Classification

多点模拟是多点地质统计学的应用,其通过模板扫描训练图像获取未知点的局部条件概率,在地物几何结构模拟上有传统地质统计学不可比拟的优越性。把多点模拟应用到遥感影像信息提取中,将结构信息和光谱信息融合到一起以提高分类精度是遥感影像分类的新思路。概述基于这种思路的CCSSM分类方法、研究进展,并对其作进一步讨论,尤其对应用CCSSM做多类别分类从理论上作可行的拓展分析。

Multiple-point simulation（MPS） is one of the applications of multiple-point statistics in geology.It can simulate the complex geometry-structure by getting the conditional probability from training image through search-tree while the traditional geostatistics has little ability to capture the complex geometry-structure.When MPS is applied into remote sensing classification,it can help us to improve the traditional-pixel-by-pixel-classification accuracy by combining the structural and spectral information.This article firstly introduces the development of new classification method which is called ＂CCSSM＂,and then discusses the revision of this methods.Furthermore,it extends CCSSM＇s application to the classification of the multiple-class from in the perspects of a theoretical and feasible way.