The Study And Application Of Face Detection And Tracking Base On Video

Face detection and tracking technology has a widespread concern in the field of computer vision. It has a very wide applications, such as content-based image and video retrieval, video conferencing, security surveillance and human-computer interaction. Also it is the basis of face recognition, face recognition and a series of more advanced applications. There are many fields that are involved in the application of face detection and tracking. This thesis aims at the problem of face detection and tracking technology. After analyzed and summarized various methods of face detection and tracking technology among the current domestic and international we choose some of the prevailing method to study.In the aspect of face detection: we choose the Adaboost as the face detection method, at the same time in order to solve the slow speed problem of face detecting when using Adaboost, we choose a combination method of Adaboost and the skin color segmentation. This combination method follows two steps to perform face detection, first do a pre-processing by using the skin color segmentation, then use the Adaboost to verify if the skin color region is a face. After using the color segmentation to filter the non-skin region, Adaboost need not traverse all the child window in the image area, it just need to detect the skin color area and thus makes the face detection more fast than before. Experiments have shown that this method is effective.In the aspect of face tracking: we choose the Cam Shift as the face tracking method, CamShift is an abbreviation of continuous adaptive Mean Shift algorithm, which is an improved version of the Mean Shift algorithm. It can be used to process video sequence. For each frame image in the video CamShift call Mean Shift algorithm to match the target. The CamShift algorithm has two characteristics: 1) CamShift use the color information of the target object when performing matching. 2) CamShift only perform local search, So when the target object moving too quickly it is very likely to lead to the loss of target tracking,Allow for such situation, we choose a combination method of CamShift and the Kalman filter. This combination method need to use the Kalman filter to predict the motion of the object, and then update the CamShift search window, So even when the target object move fast, we can still managed to track that target object, and it is unlikely to lose tracking of that object Experiments have shown that this method does achieve some good result.After studying the face detection and tracking algorithm we designed and implemented an automatic face detection and tracking system based on OpenCV(open source computer vision library).