| 摘要: |
| 用不同空间分辨率的TM与IRS－1C（PAN）遥感图像进行融合，可增强图像清晰度。本研究用人工神经网络BP算法对TM和IRS－1C（PAN）的融合图像进行土地覆盖分类，分类的总体精度达到95％，高于最大似然法（分类的总体精度为71％）。 |
| 关键词: 人工神经网络 遥感融合图像 分类 IM IRS 土地覆盖 |
| DOI： |
| 修订日期：2001-03-13 |
| 基金项目:北京市土地变更调查攻关课题资助 |
|
| Classification of Land Cover Based on Fused Image of TM with IRS |
|
|
|
| Abstract: |
| The fused product merged two optical image data of different resolutions--a high spatial resolution panchromatic image (IRS 1C) and a low spatial resolution but multispectral image (TM). Its signal clarity was improved. Artificial neural network technology is of great advantage to deal with data of uncertain distributing and qualitative data such as performing non linear classification, and thus being used to classify the land cover. The classification accuracy of fused remote sensing image reached a accuracy of up to 95%. It is far much better than the method of maximum likelihood classification, whose total accuracy is only 71%. |
| Key words: artificial neural network,fused remote sensing image,classification |
