可见-近红外高光谱成像结合化学计量学分辨人面部信息

发布时间：2018-05-14 22:23

本文选题：高光谱成像 + 化学计量学　；参考：《光谱学与光谱分析》2017年08期

【摘要】：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。
[Abstract]:Human facial information, like fingerprints and iris, can be used for identification, and it is easier to distinguish and recognize people from a long distance. Using hyperspectral imaging technology can be applied to face recognition and obtain abundant information and large amount of imaging data. It is necessary to use chemometrics to extract the effective information contained therein and to lay a foundation for computer recognition. The feasibility of visible-near infrared hyperspectral imaging to analyze human facial information is studied. The results show that the multivariate curve resolution-alternating least squares method is different from principal component analysis in that it can represent human face information through physical and chemical data of specific physical and chemical significance such as principal component spectrum and relative concentration. Moreover, it is convenient to impose the constraints in the operation according to the characteristics of the imaging data. In addition, partial least squares discriminant analysis is used to classify the spectrum of skin signals with different skin colors. The features of facial hyperspectral information of white and yellow people are similar, and the difficulty of classification is higher than that of dark skin.
【作者单位】：首都师范大学化学系;
【基金】：国家重大科学仪器专项项目(2012YQ140005)资助
【分类号】：O657.3

【相似文献】