Design And Implementation Of Facial Feature Points Localization System Based On Constrained Local Model

Recently, face recognition is widely applied in many fields. As a core factor in face recognition, facial feature points localization has received widespread attention and intensive study. It is specialized in searching for facial features such as eyes, nose, mouth and so on. It can be applied in various areas like pose estimation, facial expression analysis, three dimensional face modeling and animation, etc. So it has certain theoretical value and practical significance to study facial landmarks localization.This thesis summarizes the existing algorithms of facial feature points localization and goes into details for Constrained Local Models. This thesis analyzes the ASM and convex quadratic fitting method and improves the convex quadratic function optimization strategy. Then we achieves a facial feature points localization system based on CLM.Firstly, shape model of CLM is built by Point Distribution Model. Secondly, CLM algorithm trains a linear Support Vector Machine to get patch model. Viola and Jones' s face detection algorithm is used to detect human face and model parameters are initialized according to the face region. On the basis of the response map of SVM, the search strategies of CLM updates model parameters until facial feature points are located correctly.The results show that the improved convex quadratic optimization strategy effectively solves the problems of local minimum. So the accuracy of feature points localization is improved. This facial landmarks localization system can locate facial feature points rapidly and correctly. And the CLM algorithm outperforms the traditional ASM and AAM algorithm.