Tongue And Facial Images Fusion Analysis Research

The focus of future Medical will be shifted from treating serious illness to preventing from disease. Traditional Chinese medicine(TCM) emphasizes on prevention and combination of treatment. This characteristics of TCM are accord with the trend of future medical. Inspection has an important role in TCM diagnosis. Modernization of inspection means with the help of modern technology, TCM can obtain accurate and consistency image signal and by using the digital image and pattern recognition techniques. As the result, it becomes possible to realize the computer automatic of inspection. Tongue diagnosis and the complexion diagnosis are important parts of inspection.This thesis is a part of automation of tongue and complexion diagnosis research. It concentrates on the fusion of tongue and complexion research, including the design and manufacture of the standardization acquisition equipment, tongue and facial color feature extraction and their fusion, classification of health and diabetes based on the fusion color feature of tongue and facial.In the first part, requirements of the acquisition equipment are analyzed in detail. Analysis on four core modules, the light source, imaging camera, lighting environment and color correction are presented. Each module has to meet the demand of performance and their parameters are discussed. Then a series of tests are made to test the acquisition equipment, including illumination uniformity testing, accuracy and consistency testing. The results show that the acquisition device meets the demand well.Color feature extraction and fusion methods of tongue and face images are studied in the second part. Firstly, statistics on large-scale samples of tongue and face are done respectively to get the tongue and facial color gamut. Then, we use the fuzzy c-means clustering method(FCM) to cluster the dataset. There are 12 clustering centers of tongue and 6 of facial. At last, we fuse tongue and facial color feature vector together for the next classification.In the last part, the fusion color feature vector is using for classification. The KNN, SVM and SRC are applied to classify samples on a dataset consisting of 125 Healthy and 284 Diabetes Mellitus(DM) samples. The experimental result shows that the fusion color feature has a good performance.