TCM Complexion Recognition Based On Skin Color

Complexion and tongue diagnosis is the core of objective observation clinic.The Objectification of tongue diagnosis has been applied clinically.However,there are still much more blank areas in the Objectification of complexion diagnosis.Many researchers start to analyze the objectification techniques of TCM facial diagnosis from the perspectives of instrumentation and image processing respectively.Due to the high degree of sophistication required for the instruments and equipment in the clinic,and the high development cost of the precision equipment,the researchers focus more on the combination of TCM diagnosis and image processing.Therefore,how to more effectively use the image processing method to realize The Traditional Chinese Medicine clinic has become a highlight issue that many scholars pay close attention toThis article takes TCM facial diagnosis as the theoretical basis,based on the template and skin color model combined with the method of detection of human face and skin segmentation,and extract the facial features of the skin blocks to classify the TCM complexion.The main research contents include:(1)Propose an algorithm based on TCM face segmentation of interest in the face of the skin.Firstly.The face is initially located through template matching,and then the face in the original image is segmented by the threshold.The facial features are accurately positioned by the key points of Active Appearance Mode algorithm.Using the TCM as a guide,the forehead,left,Right eyelid,left and right cheeks,nose,mouth,jaw eight are areas of interest.(2)Proposed a skin segmentation based on the region of interest.After the located region of interest is transformed into YCbCr space,the skin patches are segmented using the elliptic skin color model and the located regions of interest are compared with the skin patch results using the skin model.The experimental results show that the segmentation results using the skin color model have higher robustness.(3)Proposed The color feature,the texture feature and the local binary pattern feature as the fusion features of the facial features of TCM and classified by the classical machine learning algorithm.Different color features were extracted from the segmented skin patches,and the normalized rgb,HSV,Lab color features,the texture features of the gray level co-occurrence matrix and the local binary pattern were taken as the final facial features.The extracted facial features are classified by using support vector machine and BP neural network,and the results are compared and analyzed.Experimental results show that the highest classification accuracy of facial fusion features is 89.608%.