The Analytic System Of Inspection Diagnosis Based On Machine Vision

Inspection diagnosis is an important part of traditional Chinese medicine(TCM) four diagnostic methods.Its definition is that diagnosing the patients by analyzing the external characteristics. In traditional inspection diagnosis, doctor relies on the experience excessively which will been influenced by environment and subjective factors easily, making the diagnosis result lacking of objectivity. So, it is very important to make inspection diagnosis objective and quantitative. Hand diagnosis and tongue diagnosis are both an important part of the inspection diagnosis. The paper was researching the objectification on hand diagnosis and tongue diagnosis, and obtained certain achievement.The objective research object of tongue diagnosis is acantha tongue and crack tongue. The following is the main content of this part, first, separating the background, face, and tongue regions according to the difference of H and I of tongue image in HSI color space. Then, eliminating burr, noise and holes using mathematical morphological, and getting the approximate outline of tongue using convex hull. Second, improving the traditional active shape model(ASM), which including that, abandoning the method of detecting the tongue area by classifier, making the approximate outline of tongue, which was got in HSI color space, as the initial model, then, searching the tongue body area and completing the tongue image segmentation. This method improved the segmentation accuracy effectively. Third, extracting the acantha and crack from the tongue image using the gray difference and extracting features from acantha and crack. Doing statistical analysis on acantha and crack features and concluding that they have statistical significance.The objective research object of hand diagnosis is palms of asthma atients and healthy people. The following is the main content of this part, first, separating the background and hand regions according to the difference of H and S of hand image in HSI color space. Then, extracting the hand from hand image combining with mathematical morphology, median filtering and outline extraction. Second, dividing the hand into interested finger region and interested palm region according to the the theory of anatomy. Extracting the interested finger region according to the position relationship between interested finger region and the fingers. Extracting the interested palm region according to the template matching. Third, Getting features from interested palm region,doing statistical analysis on palm features and concluding that they have statistical significance.Reducing the dimension of extracted features of acantha crack and interested tongue by Principle Component Analysis(PCA), Linear Discriminate Analysis(LDA), Locality Preserving Projections(LPP), getting new features. Then, classifying the features by support vector machine(SVM) and K-nearest neighbor(KNN).Last, programming the software of tongue diagnosis analysis system and hand diagnosis analysis system combing with Visual Studio2005and OpenCV.