Research On The Assistant Technology Of Tongue Diagnosis In Chinese Medicine Based On Machine Learning

Tongue diagnosis is an important diagnostic method in traditional Chinese medicine.Physicians learn about patients’ physical conditions by observing their tongues,tongue diagnosis mainly relies on physicians’ diagnostic experience,and the diagnosis results are highly subjective.In recent years,in order to collect and analyze tongue information objectively,researchers have used computer vision technology to process and analyze tongue information.Therefore,this paper puts forward the tongue body segmentation,tongue coating extraction and tongue coating classification methods:A tongue image segmentation method based on Grab Cut without interaction.In this paper,Grab Cut algorithm is used to segment the tongue body image and the interactive Grab Cut algorithm is converted to the non-interactive Grab Cut algorithm.Firstly,tongue features are used to improve Fuzzy C-means clustering algorithm,and the tongue information automatically obtained by the improved FCM replaces the interactive information needed by Grab Cut algorithm.Then,considering the influence of uneven illumination on the segmentation accuracy,an adaptive brightness adjustment method is proposed to correct the brightness of the original image.Finally,Grab Cut algorithm is used to segment the tongue body.A tongue coating extraction method based on adaptive gamma ray correction and clustering algorithm.The method proposed in this paper takes into account the small color feature distinction between tongue coating and other areas of tongue body,and uses image enhancement method to expand the distinction between tongue coating and tongue mass.Firstly,an adaptive gamma correction method is proposed to enlarge the difference between tongue coating and tongue mass in the image.Then,the K-means clustering algorithm is used to segment the enhanced tongue body image,so as to extract the accurate tongue coating.A tongue coating recognition method based on weight initialization-CNN algorithm.This paper considers that the distribution and color characteristics of tongue coating can be used as the classification standard of tongue coating.Firstly,the weight parameters and bias of the convolutional neural network model are initialized according to the color characteristics and position characteristics of tongue coating,so that the initialization model can simply represent the image features.Then,an appropriate network model is constructed to classify tongue coating,and the classification results of yellow and white coating are finally obtained.