| Through the observation of tongue to diagnose the disease of traditional Chinese medicine(TCM)inspection is a very important content,which plays an important role in the diagnosis of disease characteristics and location,as well as an important reference for the observation of the course of disease.In view of the present main tongue diagnosis in TCM based on TCM doctor’s subjective experience,lack of objective and unified evaluation criterion,by the use of digital image processing technology,many tongue features which have TCM value can be extracted and used for analysis,and the key objective parameters of TCM tongue diagnosis were obtained by further calculation of these features in this paper.By analyzing and sorting out the existing tongue image algorithms,in this paper the algorithm focuses on the tongue images’ color correction technology,the objective detection of tongue feature and the objective detection of tongue coating feature.In the part of tongue image color correction,different color correction algorithms are used according to different tongue image acquisition devices.For the tongue image acquisition instrument with high precision requirements,color card is used for color correction.For the portable tongue image acquisition instrument,the automatic white balance algorithm combining gray world algorithm and perfect reflector assumption is used for color correction.The research object of objective detection of tongue shape features is toothed tongue and tender tongue.Teeth marks in the objectification of tongue inspection of tongue body first extracted teeth marks on the edge of the area,in Douglas–Peucker point sparse algorithm was improved,on the basis of the tooth marks on the edge of the key points for extraction and analysis,obtained can assist clinical indented tongue diagnosis of traditional Chinese medicine and the number of teeth marks objective parameters,such as teeth marks,the degree of teeth marks detection accuracy reached89.63%.The objective analysis of tongue tender is to calculate the tongue tender degree and to achieve the quantitative detection of tongue tenderness through comprehensive K-means clustering method to extract the tongue body color features and its texture features which can be extracted by using gray-level co-occurrence matrix.The accuracy rate of discriminating tongue tender degree through the aging degree reaches 87.66%.The research of objective detection of tongue coating is to distinguish whether the coating has curdy or greasy properties and dry or moist properties.In order to realize the effective extraction of the characteristics of curdy or greasy tongue coating in objective detection of curdy or greasy tongue coating and give consideration to the color and texture characteristics of curdy or greasy tongue coating,the traditional SLIC super-pixel segmentation algorithm integrating Tamura roughness was used to divide the curdy or greasy tongue coating into a series of sub-images,and then the Gabor ripple algorithm was used to extracted the texture features to realize the classification and recognition of curdy or greasy tongue coating.The accuracy of the improved identification algorithm for curdy or greasy tongue coating is improved by about 30%,which is better than the original tongue coating algorithm,and the degree of greasy tongue coating is quantified by DCT transform.The objective detection of moist or dry coating is to distinguish the bright spots of water on the tongue coating from the natural areas with lighter colors through the dichotomous light model,and calculate the dryness degree of the tongue coating by the area of the bright spots of water to realize the objective description of dryness degree of tongue coating.In order to verify the tongue as a key objective parameters extraction and analysis algorithm validity in clinical diabetic retinopathy diagnosis of TCM,the tongue image data of 200 patients with different stages of diabetic retinopathy were collected for statistical analysis in this experiment,found that most patients with diabetic retinopathy had the tongues of some certain commonalities,such as toothed tongue,tongue swelling,delicate texture,thick fur and mossy quality.In addition,the degree of tooth marks,tenderness and dryness of tongue image were different with different stages of diabetic retinopathy,which verified the auxiliary role of extraction and analysis algorithm of key objective parameters of tongue image in the diagnosis and classification of diabetic retinopathy. |
