The Classification System Of Mouth Type Based On Active Shape Model

With the rapid improvement of economy and living standard, aesthetic plastic surgery are developing fast. More and more people change and beautify themselves and restore confidence through cosmetic surgery. Facial plastic surgery is a high-precision and high-risk clinical therapeutic technique, in order to improve the effect and success rate of it, the research change from cognitive psychology to computer artificial intelligence field. In recent years, the processing technology of human face has been paid more and more attention to, and has got a certain achievements, therefore it is feasible to aid medical plastic surgery by using computer technology with a considerable application prospect.Before the face surgery, accurate mouth type classification is essential. Traditional mouth type classification mainly depends the observation of doctor, which is easily influenced by subjective factors.Different doctors may obtain inconsistent classification result because of different evaluation criteria. In addition, lack of accurate measurement data also make the classification results unreliable. The mouth type classification using computer vision technology has the advantage of simple operation, uniform standards, strong promotion and so on. In this paper, the computer vision is used for the medical treatment, and the facial plastic research is quantified and automated, and it is hoped that it can be used in clinical beauty, plastic and repair.Firstly, the paper studies a facial tracking and recognition algorithm based on improved ASM. ASM is one of the classical algorithms of face feature points calibration, but the algorithm has low precision which limits its application field. The facial tracking and recognition algorithm is implemented through combining SIFT descriptors with Mahalanobis distance. It takes advantage of model-building ability of SIFT and the fast fitting efficiency of MARS, which could label facial feature with high-speed. The experimental results illustrate that this algorithm could accurately fit and track the face which has a certain angle.Secondly, from the perspective of mouth type classification, a new method is proposed for the classification of mouth. Based on the feature points of human face the algorithm set up a mouth template. After a series of operations such as rotation and binarization, the normalized feature vector is formed. The weights of BP neural network are trained by sample images. Experimental results show that the proposed algorithm can achieve more accurate classification of mouth.Finally, based on ASM algorithm the classification system of mouth type is developed by C++ and Open CV with the software platform of VS2008 and Open CV 2.4.6. The system can quickly and accurately complete facial feature points calibration and mouth type classification, the experimental results are basically in accordance with the needs of clinical medical plastic surgery. The platform is a solid foundation for further study of the eye and face classification of medical plastic surgery.