Human Face Retrieval And Event Detection Technology In Video

With the development of computing technology, the internet has becoming more and more universal, massive image and video information are stored in digital format. The retrieval of human face information is a subject with high research value. On the one hand, the human face retrieval is an important part of the content-based retrieval; on the other hand, videos provide a realistic environment to help researchers find effective way doing face recognition.Firstly, we surveyed basic technology in face detection, face feature extraction and face recognition. Three key technologies were discussed separately in different chapters. Face detection, as an example:adaboost-cascade algorithm was used to locate face region, a linear face tracker and skin model were integrated. Principal components analysis (PCA), as a traditional method, has been widely used in image processing. In this paper, PCA using in face retrieval in video also will be discussed. Fisher linear discriminating (FLD) approach was used to reduce ddimensionality further to realize the face recognition. In this paper we will do an experiment on PCA+FLD face reorganization.The proposed human face retrieval system is introduced in detail. Adaboost is used to detect human face. Face tracking is integrated to reduce the searching region. Skin model is used for searching falsely detected faces which were the results of the adaboost classifier. Experiment proved that this method can accelerate the detection speed and improve the detection rate effectively.At last, conclusions about the above approaches are drawn and future work is discussed.