基于冠层辐射传输模型的地表反射率光谱重建方法

针对基于多光谱数据有限光谱信息重建地表反射率光谱的病态求解难题，提出一种基于冠层辐射传输物理机理并充分考虑像元异质性的地表反射率光谱重建方法，该方法假设混合像元由植被和土壤两种地物类型组成，利用冠层辐射传输模型构造端元光谱查找表，进而通过组分比例因子估算实现基于多光谱图像的高光谱地表反射率模拟。以Landsat ETM+多光谱图像为例的地表反射率超光谱重建验证实验结果表明，模拟的反射率光谱能够较好的反映不同地物特征信息。进一步地，利用模拟的地表反射率拟合Landsat ETM+图像和MODIS图像，各波段模拟图像与实际观测图像之间具有较高的相关系数(Landsat: 0.90~0.99, MODIS: 0.74~0.85)，进一步验证了该方法的可行性。

A Method to Reconstruct Surface Reflectance Spectrum from Multispectral Image Based on Canopy Radiation Transfer Mo de l

Due to the lack of enough spectral bands for multi-spectral sensor, it is difficult to reconstruct surface reflectance spectrum from finite spectral information acquired by multi-spectral instrument. Here, taking into full account of the heterogeneity of pixel from remote sensing image, a method is proposed to simulate hyperspectral data from multispectral data based on canopy radiation transfer model. This method first assumes the mixed pixels contain two types of land cover, i. e. vegetation and soil. The sensitive parameters of Soil-Leaf-Canopy (SLC) model and a soil ratio factor were retrieved from multi-spectral data based on Look-Up Table (LUT) technology. Then, by combined with a soil ratio factor, all the parameters were input into the SLC model to simulate the surface reflectance spectrum from 400 to 2 400 nm. Taking Landsat Enhanced Thematic Mapper Plus (ETM+) image as reference image, the surface reflectance spectrum was simulated. The simulated reflectance spectrum revealed different feature information of different surface types. To test the performance of this method, the simulated reflectance spectrum was convolved with the Landsat ETM+ spectral response curves and Moderate Resolution Imaging Spectrometer (MODIS) spectral response curves to obtain the simulated Landsat ETM+ and MODIS image. Finally, the simulated Landsat ETM+ and MODIS images were compared with the observed Landsat ETM+ and MODIS images. The results generally showed high correction coefficients (Landsat: 0.90~0.99, MODIS: 0.74~0.85) between most simulated bands and observed bands and indicated that the simulated reflectance spectrum was well simulated and reliable.