基于偏振植被指数的植被-土壤混合像元室内实验研究

混合像元是遥感影像中普遍存在的一种现象,对其组成和各成分比例的反演一直是遥感研究中的重难点,而国内利用偏振植被指数对混合像元的研究几乎没有涉及。本次研究通过对不同面积比例的植被-土壤混合像元偏振反射高光谱特征进行分析,讨论不同条件下的植被-土壤混合像元偏振高光谱特性,并利用偏振反射比计算了12种0°偏振态下的偏振植被指数,分别构建了植被指数与植被面积比例以及光谱特征参数与植被面积比例的数学模型。结果表明,混合像元中植被面积比例和偏振角均对其偏振高光谱有一定的影响;865 nm的偏振反射比与植被占像元面积比例的相关性最好,采用多项式进行拟合时,其决定系数达到0.99,适合进行植被占像元面积比例的反演;偏振光谱“红边”处的一阶微分值与植被像元比例存在良好的线性相关,R²=0.974;植被面积比例与植被指数和光谱特征参数呈现良好的相关关系,其中P-DVI和光谱吸收指数（SAI）与植被面积比例的拟合效果最好,决定系数分别为0.99以及0.94,适合进行植被-土壤混合像元中植被面积比例的反演。

Indoor Experimental Study on Vegetation-soil Mixed Pixels based on Polarization Vegetation Index

In remote sensing images, mixed pixel is a common phenomenon. The inversion of component composition and proportion of mixed pixels has been a key and difficult point in remote sensing research. Besides, the study of the mixed pixel using polarization vegetable index has not been conducted in China, yet. This study analyzed the hyperspectral characteristics of polarized reflections of vegetation-soil mixed pixels in different area ratios and discussed the hyperspectral characteristics under different conditions. We used the polarized reflection ratios to calculate the polarization vegetation indices of 12 types of 0 degree polarization and constructed the mathematical model of vegetation area ratio related to spectral characteristic parameters and vegetation index. The results show that vegetation area ratio and polarization angle have certain influence on hyper-spectrum of polarization of mixed pixels. The correlation between polarized reflection ratio of 865nm and the proportion of the vegetation is the best. When the polynomial fitting is applied, the coefficient of determination is up to 0.99. It implies this method is suitable for the inversion of the proportion of vegetation in pixels. With the proportion of vegetation there is a good linear correlation of first order differential of the polarization spectrum on "red edge",R- square = 0.974. Spectral characteristic parameters and vegetation index have good correlations with vegetation area ratio with the P-DVI and spectral absorption index (SAI) showing the best fitting goodness of vegetation area ratio (the coefficients of determination are 0.99 and 0.94, respectively). It implies this method is suitable for the inversion of vegetation area in vegetation-soil mixed pixels.