Research On Classification Method Of Tongue Substance Color And Tongue Coating Color Based On SVM

Recently, Tongue Diagnosis has gained more and more recognition and application at home and abroad because of its simplicity and efficiency. With the rapid development of computer science and technology in these years, traditional tongue diagnosis will definitely be computerized. This dissertation is thus aimed at helping the development of computerized Tongue Diagnosis and researching on the method of color classification of tongue substance and tongue coating in Tongue images.The main contribution of this dissertation includes: uisng the framework of tongue color classification system based on pixels; proposing a method to get tongue coating and tongue substance color samples by clustering the samples after getting samples from sub-blocks of tongue images; using mutiple method of SVM to classify the 12 class of pixel color ;proposing a method used the SVM and dynamically local KNN algorithm, which improved the speed of tongue substance and tongue coating color classification; designing a decision tree according to the knowledge of traditional Chinese medicine, and applying it to the tongue substance and tongue coating color classification of tongue images.Firstly, aiming at the requirement of tongue color classification system based on pixels, this dissertation proposed a method to get tongue coating and tongue substance color samples by clustering the samples from sub-blocks of tongue images, clustering the samples can remove the impurity in the sample collection which is from the sub-blocks of tongue images. The experiment showed that the result was satisfied.Secondly, combining with the characteristics of the samples of tongue substance and tongue coating color, this dissertation used decision tree and OAO methods to solve the multi-class classification problem of tongue image pixel color classification, experiments showed that these two methods had good performance in classification; on this condition, according to the characteristics of tongue substance and tongue coating color classification, this dissertation proposing a method which used SVM and dynamically local KNN algorithm, this method was proved speedy and efficient in the classification experiment. Finally, according to the knowledge of traditional Chinese medicine, this dissertation designed a decision tree, and proposing a classification algorithm to classify the tongue substance and tongue coating color of tongue image with the 12-dimension feature vector of color ratio. The experiment result showed that the algorithm was effective.