| Face detection is a hot area of computer vision and digital video processing. Recent years face detection technology is widely used in areas such as personal identification and video surveillance, thus people more and more emphasis on the research of face detection. Face detection is the first step in face recognition, the accuracy of face detection affects directly on the accuracy of face recognition, thus face detection has important theoretical and applicable value. accuracy and efficiency are the two criterions to evaluate a face detection system.The main work of this paper is to analyze and study face detection method and face tracking method. On the face detection research, this paper presents a face detection algorithm which based on contour extraction and Haar-like feature in order to satisfy the face detection in video sequences which based on complex background. In the paper, face detection algorithm has two stages: the first stage is using the mathematical morphology and edge detection methods to extract image contour and estimating the result of contour extraction. The second stage is using face detection algorithm based on Haar-like feature to make a further estimating to the first stage result.On the face tracking research, the CAMShift algorithm mainly discussed in the paper. Because the color information is the inherent character of the object, using this character to track the moving face could solve the problem of lost caused by block. CAMShift is a algorithm based on the object color information, and it is widely used for the parameter-less and fast calculation.Finally, using the face detection and tracking system which carried out in this paper to make an experiment on simple background static pictures, complex background static pictures and image frames in video sequences. The experiment results indicate that the detection rate, false rate, missing rate and detection speed can achieve a better result. |
PDF Full Text Request