人的面部信息与指纹和虹膜一样可以用于人的身份鉴别, 并且相比之下更容易实现远距离的分辨和识别。 利用高光谱成像技术可以应用到人脸识别领域并获取丰富的信息和庞大的成像数据量, 需要采用化学计量学方法才能充分提取其中包含的有效信息, 并为计算机识别奠定基础。 研究了可见-近红外高光谱成像技术对人的面部信息进行分析的可行性。 结果表明, 多元曲线分辨-交替最小二乘方法不同于主成分分析, 能够通过主成分纯光谱和相对浓度等具有具体物理化学意义的数据表征人的面部信息, 而且可以方便地根据成像数据的特点施加运算中的约束。 另外, 采用偏最小二乘判别分析的方法实现了对不同肤色的皮肤信号光谱进行分类。 白种人和黄种人的面部高光谱信息特征相似, 分类难度高于深色皮肤人种。

The human face information, like the fingerprint and iris, can be used for the identification of a person, and is easier to achieve long-range resolution and identification. Hyperspectral imaging can be used to obtain a wealth of chemical properties and a large amount of data for human identity, while we need to use chemometric methods to extract the chemical characteristics from the image dataset, and the obtained face feature can be used for computer recognition. In this paper, we have investigated the feasibility of the analysis of human face by using hyperspectral imaging combined with chemometric methods. We compared the results of multivariate curve resolution alternating least squares (MCR-ALS) and principal component analysis (PCA). MCR-ALS gave the pure principal components spectra and their corresponding relative concentrations to display the information of the human face, and constraints could be applied conveniently based the features of imaging data. In addition, the method of partial least squares discriminant analysis (PLS-DA) was used to classify the human skin signal spectrum. From the spectra analyzed, the facial information of white and yellow people is similar, and the classification of them are more difficult than that of dark skinned people.