Research On Face Detection And Facial Feature Extraction In Video Sequences

With the increase of civil and commercial research projects, and the application requirement in safe-defence and monitoring area, human face has already become an important research object. This thesis based on the application of pattern recognition technology in intelligence monitoring area, deeply researched several pattern recognition technologies , such as, face detection, face tracking, facial feature extraction, facial expression recognition and so on. Finally, we finished a real-time intelligence recognition system.We proposed an improved face detection method, it checked the detection result of Haar-based method by the distribution of human skin colour in H channel. This improved method can delete the wrong result and improve detection efficiency. As our work also changed images to HSV Chrominance Space, this method didn't increase too much computing work.Combining with this face detection method, we proposed a face tracking method. Firstly, we detected faces in video sequences. Then we extracted colour information in these faces. At last, we tracked faces by CAMSHIFT algorithm. As the robustness of CAMSHIFT algorithm, experiment result was quite well.