Research On Key Technology In Non.marked Facial Expression Capture System

With the rapid development of motion capture technology,the system of motion capture has been widely applied in the field of films,animation production.At present,facial expressions captured by tracking light-sensitive markers pasted on actor's face has been adopted by most motion capture systems.But it is limited to facial expressions,and the marked points on face are so easily to fall out that the precise controls of performers has been reduced.Hence,it has been a new research orientation of accurate recognizing and locating facial expression by the technology of motion capture without any auxiliary marked pointsThis thesis has designed and achieved a kind of unmarked facial expression capture system.Firstly,provide a parallel face detection method basedon GPU.The system has been able to detect the initial position of face rapidly with the features of the Haar and the facial feature points combined with active shape model so that to capture 2D information of facial feature points.Secondly,with principle of binocular vision to calculate the 3D information of facial feature points and build the topological structures of facial feature points in the 3D space.Finally,process the 3D information of facial feature points with the Autodesk Motion Builder to rebuild the facial feature points and realize the real-time display of facial expression captured by driving the model of 3D facial feature points,which realized the general adaption of unmarked facial expression motion.The software of camera calibration,data collection and real-time data transmission of feature points was written by C++ language,it was interactive and easy to operate.It has been verified that the related theories and methods based on experimental simulation in this paper are with a nice application prospect.