Traditional Chinese Medicine Based Texture Analysis Of Thickness Of Tongue Coating

Recently, Tongue Diagnosis has gained more and more affirmation and application from domestic to abroad because of its simplicity and immediacy. Meanwhile, as the development of the modern science and technology, it is important and valuable to the development and application of Tongue Diagnosis to combine the computer technology such as digital image process and pattern recognition and Tongue Diagnosis in order to avoid the defect of inquantitativeness and subjectiveness of Tongue Diagnosis.Based on observations of large number of tongue images, in this thesis we determined the most obvious region of tongue where tongue-coating locates, proposed a method for sub-tongue image selection, make definite the research on objectives. Inspired by predecessors'experience and results we studied methods for extracting object tong image'textural features and designed multiple texture-feature-based tongue image classifiers. We then make a multi-classifier ensemble to enhance, and the classification performance of the ensemble was better than any single classifier.First, we chose a certain number of standard samples with different category tags, eliminate reflecting spots from each sample and use Gabor wavelet transform to extract textural featuresThen, we studied 3 different classifiers in this thesis. They are Linear differentiate, BP Neural Networks, and GA-based Optimizer.Finally, we used standard samples to train these 3 kinds of classifiers and tested their classification results on an independent test set. After that we combined the 3 single classifiers and test the ensemble's classification performance. We found that the multi-classifier ensemble got an enhanced classification performance.