摘要
与TM/ETM+相比,CBERS CCD缺少2个红外波段(波段5和波段7),这便导致了许多针对TM/ETM+数据的图像处理方法难以直接应用于CBERS CCD图像。为此,采用基于影像交叉学习的波段模拟方法,即以ETM+数据作为先验知识,通过支持向量回归(Support Vector Regression,SVR),拟合CBERS CCD与ETM+7波段DN值之间的非线性关系,进而在CBERS CCD已有波段的基础上模拟一个新的波段图像。实验结果表明,采用该方法模拟的CBERS CCD新波段与验证波段之间具有较高的相关性。
The absence of two infrared bands ( i. e. 1.55 - 1.75μm ( TM 5 ) and 2.08 - 2.35μm ( TM 7 ) ) in CBERS CCD camera compared with Landsat TM/ETM^+ results in a limitation that many algorithms developed for TM/ETM^+ images are not applicable for CBERS CCD camera data directly. In this paper, a cross - sensor image learning approach is used to simulate new Landsat - like infrared bands so as to extend spectrum coverage for CBERS CCD camera data. A support vector regression (SVR) technique is used to model nonlinear relationship between a priori knowledge from ETM^+ DN values and four CBERS CCD bands, and then new CBERS CCD bands are predicted. Experimental result shows good correlation between simulated band and corresponding ETM^+ band.
出处
《国土资源遥感》
CSCD
2011年第3期48-53,共6页
Remote Sensing for Land & Resources
关键词
波段模拟
机器学习
SVR
CBERS
CCD
TM/ETM^+
Band simulation
Machine learning
Support Vector Regression (SVR)
China- Brazil Earth Resource Satellite (CBERS) CCD
TM/ETM^+