空时互反馈人脸检测与跟踪系统的算法与实现  

Implementation of Face Detection and Tracking System Based on Space-Temporal Mutual Feedback Scheme

作　　者：李旭超[1] 边素轩[1] 刘昌鑫[1] 

机构地区：[1]井冈山大学信息科学与传媒学院,江西吉安343009

出　　处：《计算机测量与控制》2009年第6期1066-1069,共4页Computer Measurement &Control

基　　金：井冈山大学自然科学基金项目(JZ0822);江西省教育厅2009年度青年自然科学基金项目(GJJ09589)

摘　　要：针对视频序列人脸检测与跟踪算法复杂、实时性要求较强这一相互矛盾的特点,设计了基于数字媒体处理器DM642的空时互反馈人脸检测与跟踪系统;在空域,利用主成分分析和最优阈值检测人脸;在时域,根据前后帧概率密度函数的相对距离确定跟踪目标,利用卡尔曼滤波器对检测出人脸的位置进行预测,形成了空时互反馈闭环算法;实践表明,该系统运行稳定可靠,能满足实时视频监控系统的要求。In order to deal with the problem between the algorithm complexity of face detection and tracking and the strongly real- time requirement, space-temporal mutual feedback scheme of face detection and tracking algorithm based on digital media processor DM642 is proposed. Principal component analysis （PCA） and the optimization threshold is used for spatial face detection. In face temporal tracking stage, the detected face position is predicted by Kalman filter, and tracked by the relative distance of consecutive frame probability density function, which formed the close loop feedback algorithm. Testing results show that this system runs stably and accurately, which can meet real-time video surveillance and monitoring system requirement.

关 键 词：DM642 人脸检测与跟踪 空时互反馈 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]