Design And Implementation Of TCM Intelligent Diagnosis System Based On Tongue Detection And Tongue Manifestations Recognition

In recent years,with the improvement of living standards and the occurrence of the Covid-19,people begin to pay attention to their own health problems.At the same time,due to the existence of a large number of sub-health groups and some diseases that belong to the elderly are affecting the young,this makes physical examinations more normally for contemporary people,but the expensive and cumbersome physical examinations make people unable to understand their own health in time,leading to miss the optimal treatment time.In TCM tongue diagnosis,you can get a preliminary understanding the body’s the current state only by observing the appearance of the tongue,thereby giving an early warning.However,TCM tongue diagnosis is too subjective and poor in timeliness.In order to solve the above problems,we use image recognition technology based on deep learning and Internet-related technologies to develop an Android-based intelligent TCM diagnosis system,in which tongue diagnosis function is achieved based on tongue detection algorithm and tongue manifestations recognition algorithm.The voice questioning function is added to improve the accuracy of diagnosis.The health management module is added to provide users with the function of disease management,including recommending medicated meals and early warning of the disease.The personal center module and the background management module are added to improve the system information management function and ensure the ease of using and maintainability of the system.In this way,users can check their own physical conditions anytime and anywhere with mobile phones,and treat timely based on the system’s diagnosis.In our work,the YOLOv3 algorithm and the attention mechanism are combined into the Attention-YOLOv3 algorithm to realize the tongue detection algorithm,and the AdaBoost algorithm and CNN are combined into the AdaBoost-CNN algorithm to realize the tongue manifestations recognition algorithm.At the same time,the tongue detection algorithm and the tongue manifestations recognition algorithm realized the diagnosis function of user’s tongue.To a certain extent,it realized the objectivity and timeliness of tongue diagnosis,and realized the online deployment of algorithms with Flask.For the above algorithms,we have the following innovations:(1)In order to reduce the impact of background on tongue detection and improve the accuracy of the border,the attention model CBAM and SE modules are added to Dark Net-53 which is the YOLOv3 backbone network.(2)In order to improve the speed and accuracy of the tongue manifestations recognition algorithm,the Ghost Module is first used Instead of ordinary convolution to reduce the amount of calculation,and the AdaBoost in the boosting algorithm is used to enhance the CNN tongue manifestations recognition model.In the experiment,the Attention-YOLOv3 algorithm reach a m AP value of 91.41% in the case of a small sample,and the AdaBoost-CNN algorithm reach an accuracy rate of 93.35%in the same case.At present,the development of the first version of the TCM intelligent diagnosis system has been completed,which can perform preliminary tongue diagnosis and health management for users to meet the basic needs.