Fast Tongue Image Segmentation Algorithm Based On Convolutional Neural Network And Tongue Diagnosis

Traditional Chinese medicine(TCM)is one of the quintessence of traditional Chinese culture.Based on the conception of people oriented,TCM is becoming more and more popular.And tongue diagnosis is an important part in the “inspection” of TCM.But the traditional tongue diagnosis results are dependent on subjective judgments of the doctor,and the diagnosis results will be disturbed by the external environment.Automated tongue image diagnosis system that combines TCM theory and imaging technology in Western medicine is the development direction of modern TCM tongue diagnosis.In this thesis we focus on the tongue image segmentation and the color features and the teeth-printed tongue features.The accurate separation of tongue body from tongue image is the premise of recognition and diagnosis.Most of the proposed tongue segmentation algorithms are based on the improvement of traditional approaches.These algorithms can improve the segmentation accuracy of tongue image to some extent,but they are less robust.To address above problems,a method of fast tongue image segmentation algorithm using convolutional neural network is proposed in this thesis.The network is inspired from U-net which needs small amount of data,and group convolution and channel shuffle are used to improve the segmentation speed.The HIS color space is used to extract color features.what's more,the body and coating of tongue separation method and weighted average algorithm are used to reduce the influence of the coating in color feature extraction.Finally,the circularity and convex hull are used to analyze the teeth-printed.The experimental results show that the improved tongue image segmentation algorithm can improve the segmentation speed and guarantee the segmentation accuracy.What's more,it is more robust than traditional segmentation algorithms.Based on the tongue image segmentation,the color features of the tongue are extracted.The results show that the improved color feature extraction algorithm can be used to classify tongue images.The results of teeth-printed features extraction show that the circularity and convex hull can be used to classify the teeth-printed image and normal tongue image.