基于改进线性光谱分离模型的植被覆盖度反演

线性光谱分离技术可以有效地提取像元水平上植被或其他端元(影像中的地物)的相对百分比,但是目前该技术在多光谱宽波段影像数据应用中,由于波段数量、波段宽度等的限制,估算精度离定量研究的水平仍有一定差距。鉴于此,本文提出了一种改进的线性光谱分离方法,该方法在对影像进行土地覆盖分类基础上进行分离,一方面同类土地覆盖类型内同种地物的光谱变异相对较小,更有利于端元选取;另一方面,分影像的地物种类数量明显少于整幅影像,更容易满足模型的适用条件,从而突破了波段数量限制,同时使地物光谱分离更具针对性,经过验证,该方法较传统分离方法相比植被覆盖度的反演精度可提高6.4%,用该方法实现了研究区的植被覆盖度的定量反演并对研究区植被覆盖度的空间结构进行了分析。

Inversion of Canopy Abundance Based on Improved Linear Spectral Unmixing Model

Linear Spectral Unmixing(LSU) could extract endmembers such as canopy or other objects at pixel level,but the accuracy of LSU which is presently used in multispectral and relatively broad spectral range data is not available to quantitative research,so an improved technique for LSU which is based on the classification of land cover was employed in this study.First the ASTER image covered the study area was spatially segmented by five types of land cover maps.Sequentially based on the endmembers selection procedure the LSU was applied to each sub-ASTER image and full image respectively for subsequent comparative analysis.Because the number of objects in sub-image was less and the spectral variance of the same objects was smaller than the full image which is commonly used in traditional LSU,so this improved technique could break the limits of the traditional multispectral data which has less bands for hyperspectral analysis,and also,the mixed objects which occur in traditional LSU algorithm could be unmixed effectively by the new method.We concentrated our study on site of Fuzhou in Fujian Province equipped with ASTER data set.Only the first nine bands in VNIR and SWIR of ASTER were selected for subsequent analysis because the five TIR bands were not relevant to the reflectance of land surface objects. The result proved that an improved inversion accuracy of canopy abundance of-6.4% was achieved comparing with traditional Linear Spectral Unmixing(LSU).Furthermore,following the inversion of canopy abundance using the improved method,density slice was applied to the canopy abundance image and the spatial distribution of canopy abundance was subsequently analyzed.