目的：探讨中医面诊中光泽信息客观识别的方法。方法：结合计算机视觉，利用计算机辅助进行面部光泽判断，尝试将偏最小二乘法（PLS）和线性判别式分析（LDA）方法在4 种不同色彩空间下进行实验，做为面部光泽信息提取的手段。结果：PLS、LDA、2DLDA 在RGB、HSV、Lab 这些3 通道的色彩空间上的判断正确率均高于单通道的判断结果；不同的特征抽取方法在不同色彩通道上得到的正确率不同：PLS 方法在Lab 颜色空间上对人脸光泽的判断正确率为89.06%，LDA 在Lab 颜色空间上判断正确率为88.69%，2DLDA 在RGB 颜色空间上判断正确率为89.00%。结论：不同特征抽取方法对于识别中医面诊光泽信息都具有积极作用，为中医望诊中光泽的量化检测技术研究提供了一种新的方法和思路。

The study was aimed to explore objective recognition method for gloss of facial complexion in traditional Chinese medicine (TCM). Computer vision skills were utilized and feature extraction methods, such as PLS, LDA and 2DLDA, were applied to face samples in 13 color spaces as facial gloss extraction methods. The results showed that PLS, LDA, 2DLDA in RGB, HSV, Lab color space of these three channels were higher than on which is right to judge the results of single-channel. Feature extraction methods in different color channels are different from the correct rate. PLS may reach to the rate of 89.06% in the Lab color space. LDA may reach to the rate of 88.69% in the Lab color space. The 2DLDA may reach to the rate of 89.00% in the RGB color space. It is concluded that these three methods have a positive effect on gloss information extraction. It provides a new method for TCM observation of gloss quantification examination.

国家自然基金项目（30600796）：基于多种信息处理技术的面色诊信息自动识别研究，负责人：李福凤；校级杏林学者计划项目（R110203）：中医面诊多特征信息提取与融合研究，负责人：李福凤；上海市教委重点学科（第三期）（S30302）：中医诊断学，负责人：王忆勤。