The Design Of Face Dynamic Detection And Real-Time Tracking System

With the development of information technology and the requirement of applications,face detection and tracking have attracted numerous researchers.Various methods of face detection and tracking appear and related productions come out continuously.Firstly, this paper introduces a designing method of human face dynamic detection and real-time tracking system. For saving system's development time, development tools will be selected reasonably.Secondly, in order to satisfy the system's real-time requirement, the human face detection algorithm based on AdaBoost is adopted. This algorithm's training time too long and can not satisfy real-time requirement for bigger size image. According to the sequence image and human face's characteristics, the training time is reduced by selecting rectangle module reasonably and optimizing rectangle characters. Frame-difference method is introduced, so the human face detection speed is improved by reducing human face detection region area.Thirdly, compared with several familiar algorithms of human face tracking, according to the system's real-time requirement, the CamShift algorithm is adopted. This algorithm can't implement automatic follow,besides will be interfered by complexion background and shading objects.The combination of area and proportion threshold value can resolve above disadvantage. This algorithm's anti-jamming is improved through introducing brightness and saturation information and adding face weight.In addition, this paper implements the system's video capture, processing, compression, saving and rendering based on DirectShow technology. The video images can be compressed by many methods.Finally, a human face dynamic detection and tracking system is accomplished, which can be put into real surveillance project. In the process of tracking, this system adjusts manipulating console and the lens's parameters automatically according to the system's feedback information, which makes the tracking effect best.