基于点分布模型的可变形手势跟踪方法  

Tracking of Deformable Hand Gesture Based on Point Distribution Model

作　　者：赵永嘉[1] 戴树岭[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《系统仿真学报》2009年第S1期1-5,共5页Journal of System Simulation

基　　金：国家863基金(2009AA01Z333)

摘　　要：基于单目视觉的可变形手势跟踪是采用人手作为输入的人机交互系统的重要问题。提出了一种基于点分布模型的可变形手势跟踪方法。首先结合背景消减以及基于新的色彩空间的肤色检测技术对手进行检测,引进等效圆特征对静态手势进行判别,用点分布模型描述手形变化,并提出了一种手形主特征的估计方法,实现了可变形手势的自动初始匹配。然后基于卡尔曼滤波对感兴趣的手形特征点进行预测和更新。实验表明,该方法适用于稳定的、实时的可变形手势跟踪应用。Tracking deformable hand shape based on monocular vision is an important issue in gesture-based interaction. A novel method including two main steps based on point distribution model is presented. Firstly, auto-matching predefined posture with image features is achieved by estimation for the dominant feature component. A trained color coordinate is employed in skin color detection for locating hand region after background subtraction. Then, features derived from equivalent circle are used for posture recognition and point distribution models to describe deformable hand shape. as the second step, feature points of interest in hand contour are predicted and updated by kalman filter. Experiments show that the method is applicable to the stable, real-time deformable gesture tracking applications.

关 键 词：手跟踪 点分布模型 主动轮廓模型 人机交互 卡尔曼滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]