MNF和SVM在遥感影像计算机分类中的应用  

Application of MNF and SVM in Classificationof Remote Sensed Image

作　　者：纪娜[1,2] 李锐[1,3] 李静[4] 

机构地区：[1]西北农林科技大学资源环境学院,陕西杨凌712100 [2]杨凌职业技术学院,陕西杨凌712100 [3]中国科学院水利部西北水土保持所,陕西杨凌712100 [4]武警工程学院,陕西西安710086

出　　处：《水土保持通报》2009年第6期153-158,共6页Bulletin of Soil and Water Conservation

基　　金：国家重点基础研究发展计划(2007CB407203)

摘　　要：由于黄土高原地形复杂,单纯采用监督分离变换MNF(Minimum Noise Fraction)变换得到的4个去除噪声波段、归一化植督分类方法很难获得理想的精度,以延安市区为实验区,以TM遥感图像的最小噪声被指数NDVI和该地域的DEM作为数据源,采用支持向量机SVM(Support Vector Machine)的方法对研究区土地利用与覆盖状况进行分类,获得了较理想的分类结果。The classification accuracy is unsatisfactory in the complicated terrain area of the Loess Plateau when the single supervised classification is used in remote sensing. The paper discusses the extraction of clas- sification information of Yan＇an City and nearby area from a TM image and deals with the image classifica- tion based on the SVM method integrating the information of MNF, NDVI, and DEM. In comparison with Maximum Likelihood and SVM method of single spectrum, results showed that the objects with the same spectrum are distinguished by using DEM in image classification. Compared with the traditional classification method, the classification based on the information of DEM and multiple bands supported with the SVM method can acquire higher classification effect.

关 键 词：最小噪声分离变换 支持向量机 黄土高原 遥感图像分类 

分 类 号：TP97[自动化与计算机技术]