The Technique Of Face Detection And Tracking In Video Sequences

Many advanced intelligent video surveillance technology are involved in currentresearch projects, using the pattern recognition, image processing, image analysis andcomputer vision on target detection, tracking and recognition in the real-time scene. Andthe certain degree of analysis and timely response of the target play an important role ininformation security, security protection, intelligent traffic managment. Therefore,effective automatic target detection and tracking in video sequences becomes the basic ofvideo surveillance technology, and it has drawn more and more attention of researchers.This paper mainly studies and analyses the face detection and tracking technology invideo sequences. Specific content focuses on face detection algorithm based on Adaboostand face tracking algorithm based on Mean Shift. In the face detection process, skin-colorsegmentation is used, non-face regions are fast excluded, face regions are located roughly,and then uses the classifier trained by the Adaboost algorithm for further detects toimprove the efficiency of detection. In the subsequent process of face tracking, colorhistogram features in face regions are extracted to build a face template, then a certaineffect of tracking can be achieved by using Mean Shift algorithm to track. At the time, theface LBP(Local Binary Patterns) texture feature is added. Using face local colorhistogram and LBP texture features as the target feature for tracking, can inhibit the effectof smooth background and noise points on the target, tracking performance is also betterin some complex background.Based on the above content, the detection and tracking system of faces in videosequences implemented by the use of open-source computer vision library OPENCV hasgood performance and practical value, can meet the basic requirements of videomonitoring system in general environment.