Research On Feature Extraction For Face Color And Shape Classification For TCM Observation

Under the guidance of "strategic plans for development of traditional Chinese medicine(2016-2030)" by the state council,in recent years,the information technology research of our country traditional Chinese medicine(TCM)is developing,and the objectification and intelligent study of TCM four diagnostic also cause wide attention.TCM observation is to diagnose patients' illness by observing patients' face,which is the necessary job for clinical diagnosis and relies heavily on doctors' subjective qualitative diagnosis.This paper combines the computer vision technology with TCM observation theory and makes objective quantitative research on TCM observation by information technology.The content of this article is surrounded by two aspects in the field of TCM observation,face color and shape feature extraction,there are two main jobs:(1)This article puts forward a kind of face color block-mean feature extraction method based on multiple color space combination for face classification.extracting face color feature of HSI and Lab color space model from segmented face blocks,then training support vector machine(S VM)based on radial basis function kernel to predict the color category of input images.The experiment result shows:The block-mean feature extraction method is better than the row-column-mean on the basis of combination of HSI and Lab color space model.(2)This article puts forward a kind of face shape feature extraction method based on the histogram of oriented gradients and the skeletonization shape proportion for face shape classification.The input images is pretreated by graying and organs covering in order to prevent the side-effects on face shape feature extraction from unrelated interference information.Using Otsu optimal global threshold processing algorithm to segment image foreground and background,and defining a proportion features representation method based on the skeletonization algorithm,which represents face shape features from image shape perspective.The experiment result shows:The feature extraction based on the histogram of oriented gradients and the skeletonization shape proportion has a well performance for face shape classification.