Classification of Tongue Color Based on Convolution Neural Network

The color of the tongue is closely related to the patient’s physical condition and pathological condition, and it plays a very important position in the treatment and diagnosis of traditional Chinese medicine (TCM) clinical medicine. Convolutional neural network (CNN) has achieved fruitful results in image classification. Therefore, the method combining CNN with tongue color classification is proposed. Frist, initial data set is obtained by standardizing tongue image acquisition and tongue image preprocessing. Then, the Otsu method is applied on the image multi-channel to remove the background of the tongue accurately as the input of the model. At the same time, data augmentation is applied to avoid over-fitting of the model and improve the accuracy of model classification. The accuracy of the trained tongue color classifier based on CNN is 90.5% in the clinically collected data set, which is better than the traditional machine learning methods for tongue color classification.