The Research Of Tongue Feature In Two Deficiency Of Diabetes Based On Digital Technology

With the increasing quality of life,the diabetes has become a common and frequently occurring disease in today's society.The diabetes has been treated according to Chinese medicine four diagnostic methods from the overall with the advantages that the western medicine can't replace.Tongue inspection is the key part of traditional Chinese medicine(TCM),and becomes an important research direction of medical diagnosis due to simple,painless,noninvasive characteristics.As the traditional tongue inspection are observed mainly by the doctor's naked eyes,making the diagnosis a strong subjectivity.In order to solve the problems existing in the objective standardization of tongue inspection and reflect the theory of TCM treatment based on syndrome differentiation,this paper chooses the typical syndromes of type 2 diabetes traditional Chinese medicine——Qi and Yin deficiency and blood stasis syndrome as the research object,discusses the digital image processing technology in the study of the objective standardization of tongue inspection from the aspects of tongue segmentation,separation of tongue body area and tongue coating area,tongue feature extraction and recognition.The following aspects of the work are mainly done:Firstly,in order to accurately segment the tongue from the background area of the face,a tongue segmentation algorithm based on HSV color space is proposed.According to the difference of the brightness information between the tongue and the non-tongue area,the tongue image is transformed into the HSV space,a complete tongue area is obtained according to the divided the tongue image by H component and V component.Then,according to the difference of tongue body area and tongue coating area in the color information,the tongue image is transformed into Lab color space while a kind of tongue body and tongue coating separation algorithm based on K-means clustering is proposed.At the same time,the K-means clustering segmentation experiment is carried out under different color spaces and the tongue body area and tongue coating area can be a good separation.Secondly,using the color space model transformation,the separated tongue body and tongue coating area are quantified and analyzed in three color modes of RGB,HSV and Lab.The representative and high degree of distinction combination of tongue body and tongue coating color characteristics are chosen to combine with clinical characteristics of type 2 diabetes.The crack region is extracted by a fast detection crack region algorithm,and then tongue crack features are determined by calculating the regional contrast index and the regional consistency index.Finally,after the color feature of tongue body and tongue coating and the feature of crack are extracted,the tongue of type 2 diabetes with Qi and Yin deficiency and blood stasis syndrome and normal tongue are classified and identified by the K-Nearest Neighbor classification algorithm and the Support Vector Machine(SVM)algorithm.The sample data were grouped and the evaluation results were used to evaluate the performance of the classifier.The experimental results show that the comprehensive performance of support vector machine is higher than that of K-nearest neighbor algorithm and the recognition rate reached a maximum of 92.59%,which proves the validity of the feature selection and achieves the purpose of identification based on the characteristics of syndromes.