一种基于无人机高光谱数据的植被盖度估算新方法

从分析对植被覆盖度(FVC)敏感的光谱特性入手，使用Avafield-3光谱仪(测量范围300～2 500 nm)，利用人工草坪控制植被覆盖度的方式研究混合光谱与植被覆盖度的关系，通过实验发现红边区间(680～760 nm)对植被覆盖度最为敏感，而红边区间光谱的一阶导数与植被覆盖度的相关性最高(>0.98)，且有较强的稳定性，因此选择红边斜率k作为估算植被盖度的参数。参考混合光谱分解法反演植被覆盖度的经典模型--即以NDVI(normalized difference vegetation index)为参数的植被覆盖度反演模型，以红边斜率代替NDVI构建了2个反演植被覆盖度FVC的新的红边斜率模型，该模型是对经典模型的进一步改进。为验证模型精度，以研究区内无人机(UVA)的高光谱数据和研究区实际测量的植被覆盖度数据进行验证：对高光谱数据计算每个像元680～760 nm之间的斜率，利用PPI(pixel purity index)提取纯像元，计算纯植被像元光谱斜率的最大值和纯土壤像元光谱斜率最小值，利用新的红边斜率FVC模型求取植被覆盖度；实测数据采用照相方法，经过几何校正、监督分类后统计植被覆盖度，结果表明：通过实测数据与无人机高光谱数据获取的植被覆盖数据进行验证，新构建的基于红边斜率的两个植被覆盖度模型的精度(R2分别达0.893 3和0.892 7)都略高于以NDVI为参数的模型(R2分别达0.839 9和0.829 9)。提出使用红边斜率计算植被覆盖度的模型，具有较明确的生物物理意义，具有较高的应用潜力和推广价值。

A New Method for Estimating the Fractional Vegetation Cover Based on UVA Hyperspectrum

This paper analyzed the spectrum characteristic which is sensitive to the fractional vegetation cover (FVC).The red-edge slope(k) was set as the parameter of the FVC estimation model in the study.The relationship between vegetation cover and mixed spectrum was studied by controlling vegetation coverage of lawn with avafield-3 spectrometer (measuring range 300 ～2500 nm).The result showed that the red edge region(680～760 nm)was most sensitive to the fractional vegetation cover and the correlation between the first derivative of red edge region's spectrum and fractional vegetation cover was the highest (>0.98)which was steady at the same time.By referring to spectral misture analysis method for the classical inversion FVC model using NDVI as the parameter of the FVC estimation model,this paper established two new inversion models using red-edge slope instead of NDVI,improving the classical model.The accuracy of the models was verified by experiment using UVA hyperspec-tral data and vegetation coverage data measured in the study area.We calculated the slope between 680～760 nm of each pixel in hyperspectral image,extracted pure pixels by PPI,calculated the maximum spectral slope value of pure vegetational pixel and the minimum spectral slope value of pure soil pixel,and assessed the FVC by the two new models.The FVC of measured data were calculated by the method of photography after geometric correction and supervised classification.The result of the fitting analysis showed that the accuracy of two new red-edge slope models (R 2 =0.8933,0.8927)were higher than the NDVI model (R 2 =0.8399,0.8299).This model has higher physical and biologic meanings,application potentiality and promotion value.