基于混合像元分解的喀斯特石漠化地物丰度估测

我国西南喀斯特地区长期存在以石漠化为特征的土地退化问题,是我国三大生态问题之一。喀斯特地区地表复杂度高,具有高度时空异质性,像元混合现象严重,植被、裸岩和裸土为喀斯特地区典型地物,使得评价喀斯特石漠化的关键指标(如裸岩率、植被覆盖度)获取比较困难,高光谱遥感在混合像元分解方面有独特优势,可以获取地物端元的丰度。通过地面试验表明光谱指数能够表征地物覆盖度,进而以Hyperion高光谱影像为数据源,利用连续最大角凸锥方法从影像中提取这3类地物的端元,运用半约束和全约束线性光谱分解方法估算其丰度。研究表明:半约束线性分解得到的丰度优于全约束分解结果,其反演的植被、裸土和裸岩的丰度与相应的光谱指数间具有显著线性相关性,确定系数R²分别为0.92、0.66与0.84,表明地物丰度能够表征其覆盖度。因此,通过混合像元分解算法反演地物丰度来提取喀斯特石漠化因子具有一定的可行性,这为高光谱遥感在喀斯特石漠化中的评价和监测奠定了理论和算法基础。

Estimating Karst Rocky Desertification Feature Abundance by Pixel Unmixing

In recent years,land degradation characterized by rocky desertification in karst areas of southwestern China becomes one of China’s three major ecological problems.Karst areas are areas of high complexity surface,and with high spatial and temporal heterogeneity and seriously mixed pixels.Vegetation,bare rock,bare soil are typical features of Karst regions.Thus it makes it more difficult to extract the key indicators like the fractional cover of vegetation and exposed bedrock which is required to evaluate how serious the rocky desertification is.Hyperspectral remote sensing has unique advantages on unmixing,and can get the abundance of surface features endmember.First,ground tests showed the spectral indices could characterize the feature coverage.Secondly,based on the data of Hyperion hyperspectral images,the study proposed endmember extraction of three types of classes feature from Hyperspectral Images,and estimate abundance by semi\|constrained and fully constrained linear spectral decomposition method.Results showed that:semi\|constrained linear decomposition method was better than fully constrained,and its abundance of inversed vegetation,bare soil,bare rock was much more suited for spectral indices.The proposed feature abundance was able to characterize its coverage with R²0.92,0.66 and 0.84,respectively.The linear unmixing method inversed the feature abundance to extract the Karst indicator was feasible.This study indicates that hyperspectral remote sensing laid the foundation for the karst rocky desertification assessment and monitoring.