高光谱遥感技术对文物完全无损， 适于对中国古画等文物进行颜料的鉴别与分析， 但是中国古画中混合颜料不同成分的定量分析仍然是文物颜料分析领域中的难点。 针对中国古画中经常出现的混合颜料现象， 以石青、 石绿两种典型的矿物颜料为例， 选用粒度大小相同的两种矿物颜料粉末， 按照不同的体积比例精确配比获取颜料样本， 然后精确控制实验条件获取其光谱。 对混合光谱分别用全约束最小二乘法进行全波段光谱解混， 用比值导数解混算法进行单波段光谱解混， 然后评定解混精度并对解混结果进行对比分析， 探讨这两种矿物颜料的光谱混合模型。 实验结果表明， 石青、 石绿两种矿物颜料的光谱混合从总体上来说非线性混合特征较强， 但是在局部某些波段又基本符合线性混合模型， 利用这些波段采用比值导数法解混， 可以得到远高于全波段解混精度的定量分析结果。

Hyperspectral remote sensing technology is completely non-invasive for cultural relics, and suitable for identification and analysis of pigments of Chinese ancient paintings and other cultural relics, but the quantitative analysis of the mixed pigments composition in the ancient paintings is still a difficulty. For the mixed pigments phenomenon which often appears in Chinese ancient paintings, taking an example of azurite and malachite, two typical mineral pigments, we choose the two kinds of mineral pigment powder with the same size, precisely compound these two kinds of pigments to obtain pigment samples, and then obtain their spectra understrict control of experimental conditions . For mixed spectra, we use fully constrained least square method for spectral unmixing with full bands and use derivative of ratio spectroscopy for spectral unmixing with single band, then evaluate the unmixing accuracy, compare and analyze the unmixing results, and finally discuss the spectral mixing model of these two kinds of mineral pigments. Experimental results show that the spectral mixtures of azurite and malachite display strong nonlinear mixing characteristics overall, but are in accordance with linear mixing model in some strong linear bands. Using derivative of ratio method for spectral unmixing at these bands, we can achieve much higher unmixing accuracy than spectral unmixing with full bands.