光子学报, 2009, 38 (9): 2381, 网络出版: 2010-05-10
用于下垫面分类的卫星数据多波段优化选择
Wavelength Band Selection Method of Satellite Data for Underlying Surface Classification
物理光学 波段选择 正交子空间投影 Physical optics Wavelength band selection Orthogonal subspace projection MODTRAN MODTRAN
摘要
在研究用于卫星影像下垫面分类的三波段优化选择算法基础上, 探索了用于地物分类的多波段优化选择方法.通过分析目标谱分布, 并运用MODTRAN模拟计算进入卫星探测器的背景程辐射和沿卫星观测方向的整层大气透过率, 得到对应不同目标的权函数.根据权函数谱分布以及波段指数并结合卫星波段用途, 最终得到用于下垫面分类的卫星多波段优化选择结果.对结果进行分类验证, 发现此算法用于下垫面分类具有较好的效果.
Abstract
Based on three bands optimal selection algorithm for underlying surface classification, the multiband selection method to select satellite data optimal bands for underlying surface classification is studied. The weight functions are obtained by analyzing target spectra distribution and background radiance received by satellite sensor, transmittance along the satellite observation computed by MODTRAN. Satellite optimal bands for underlying surface classification are selected by using weight function spectra distribution, band index and the usages bands of MODIS. Classification result shows that the algorithm is effective for underlying surface classification.
范伟, 吴文玉, 何彬方. 用于下垫面分类的卫星数据多波段优化选择[J]. 光子学报, 2009, 38(9): 2381. FAN Wei, WU Wen-yu, HE Bin-fang. Wavelength Band Selection Method of Satellite Data for Underlying Surface Classification[J]. ACTA PHOTONICA SINICA, 2009, 38(9): 2381.