基于肤色模型与粒子滤波的人脸自动检测、跟踪算法  

An Automatic Face Detecting and Tracking Algorithm Based on Skin-Color Model and Particle Filter

作　　者：张割[1] 马乔兵[1] 孙海鹏 

机构地区：[1]北京航空技术研究中心 [2]94637部队

出　　处：《军民两用技术与产品》2011年第10期57-59,共3页Dual Use Technologies & Products

摘　　要：首先通过一种基于亮度聚类的肤色模型,检测人脸的初始位置;其次提出了一种基于粒子滤波的特征融合跟踪算法,用颜色直方图和方向梯度直方图来描述目标,在粒子滤波框架下将目标颜色和梯度信息有机结合,并自适应更新来跟踪人脸位置。实验表明,该算法不仅提高了跟踪精度,而且具有较强的鲁棒性。The position of faces was firstly detected by a skin-color model based on brightness-clustering, and then a multi-cues fusing tracking algorithm based on particle filter was presented, in which objects were represented by histogram of color and histogram of oriented gradients. Color and gradient information are fused and updated online for face-tracking. Results of experiments show that the algorithm is more precise and robust.

关 键 词：人脸检测 粒子滤波 方向梯度直方图 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]