Algorithm For Auto Face Detection And Tracking Based On Mean Shift

Face detection and tracking is an important research aspect in artificial intelligence and computer vision.As a key technology of face information processing,it has a broad application values in many fields such as video surveillance,content-based image retrieval,video conference,etc.In this thesis,a study on research and application of face detection and tracking is presented.Firstly,we use the algorithm based on face segmentation and SVM to detect the faces.Secondly,we use the mean shift algorithm to track the faces.We realize automatic face tracking of image sequence eventually.The main research work and innovation are as follows:(1)Research the face segmentation method based on face color,using YCbCr color model for colour segmentation.In YCbCr color model,we can remove the impact of image brightness.Experiments show that color segmentation method can remove large number of background quickly,then we could get the initial face candidate region so that the follow-up of face detection module can be calculated on the relatively small area,so as to significantly improve the implementation of thealgorithm efficiency.(2)Propose a face detection algorithm based on face color and SVM.The classification of SVM is a new method of statistical study which has advantages with high efficiency,low rate of wrong identification.After color segmentation,the large number of background information has been removed,greatly reduce the workload of SVM.It's a good face detection method and experiments proof the method is very efficent.ive.(3)Descipt and analyze the tracking algorithm based on mean shift.Propose a algorithm of automatic face tracking based on mean shift which use mean shift algorithm to track the face detected.To ensure the autimatic face tracking,we add face sequence to preserve faces.If a new face found,add it to the sequence,and use mean shift algorithm to track in the next frame.Experiments show that the algorithm is rapid and reliable.Finally,sum up the thesis,analyse the current research in the need to be further improved,pointed out the direction for future work.