The Research Of Face Detection And Tracking In Color Image

Face detection ,which is the pre-work of face recognition, has the application in a wide range of fields such as personal identification, visual surveillance and content-based retrieval. Human-face concerned applications is getting nearer and nearer to practical use, thus resulting in more and more emphasis over research for human-face detection. Imthis dissertation, we have made a deep study on automatic face detection and tracking in video surveillance system.The research of face detection advanced from simple image process to complex real-time video process. At present, the main methods of human face tracking are summarized as follows: Based on skin color information, model, movement information and partial organ characteristic and so on several kinds. However, face detection is a most challenging task because of the complicated pattern and the frangibility of human face, most methods have the weakness of large computation, low efficiency and many false reports among the detection result;Human skin color has been proven to be an effective feature and widely used in face detection in resent years. The results of experiment show that this detection approach features a rapid inspecting speed and is insensitive to the posture. But how to improve the performance of the skin detector is a challenging problem.On the basis of a deep study on automatic face detection and tracking in color image sequences, this paper systematically presents the previous work relative to it and introduces a novel multi-stage face detection algorithm, which makes a good use of human face pattern's valuable information in colour image sequences and subdivides the complicated detection task into three steps: the preprocessing, which aims to gain skin-colored regions with human skin color model; the roughly detection and face region refined by elliptic curve fitting; the fine detection with facial features' detection and location. First we discuss the skin's performance in difference chrominance space in the discussion of skin color model. During the Statistics of some sample skin we can build the skin color model and using the morphological process select the candidate region. Then the candidate region can be refined by elliptic curve fitting method. Further in the thesis we bring forward a modified deformable template for eyes detection and get the mouth region by making use of the mouth's performance in color space., it also be proved in the exam and represent a good result. Experimental results demonstrate that the detection algorithm is efficient in complex background. To achieve the successful tracking,besides the dependence determined and confirms a match criterion, the search space criterion brings into play a role, specially the human face which unceasingly changes the size. The movement prediction may effectively reduce the searching region, and enhances the track speed. The original tracking methods mostly use the overall searching to carry on the tracking task, which greatly increase the spending, and reduce the real-time effect.This article uses the movement characteristic of human face in the picture frame, and provide state prediction including speed and acceleration, at the region in which the human face possibly appears. Finally, this paper also proposes a new method for face tracking. The method is based on Kalman filter which combines both the motion estimation technique and the adaptive template matching technique. And this face tracking system enables the robust tracking of face with translation, zoom in/out. In the end of the thesis, bring forward some suggestions about how to expand the detection system and the follow-up operation as face recognition.