摘要
针对手势视频序列中由于手势快速运动导致跟踪失败的问题,提出一种基于mean shift的线性预测方法以实现图像序列中手势实时跟踪.该方法通过提取手势肤色HSV空间中的H分量建立目标模型对运动手势进行跟踪,并针对手势快速运动的情况利用线性预测方法对下一帧中手势的起始中心进行预测,同时更新手势的目标模型以适应光照等环境的变化.实验结果表明:在手势快速运动时该方法可对目标起始中心进行有效预测,提高手势跟踪的精确度.
Aimed at the problem of hand gesture tracking failure due to quick motion of its video stream,a linear prediction algorithm based on mean shift was proposed to track real-time hand gesture.This algorithm was used to abstract H component of skin-color information in HSV color space and build gesture model for tracking.When hand gesture moved drastically fast,this algorithm would predict object＇s next frame beginning center,and update gesture＇s model to adapt to alternating environment.Experiment result proved that,this algorithm could effectively predict gesture＇s beginning center and improve hand gesture tracking precision greatly.
出处
《兰州理工大学学报》
CAS
北大核心
2010年第2期75-78,共4页
Journal of Lanzhou University of Technology
基金
国家自然科学基金(60572011)
甘肃省教育厅硕导科研基金(0703-08)
关键词
线性预测
手势肤色
手势实时跟踪
linear prediction
skin-color information
real-time hand gesture tracking
作者简介
张秋余（1966-），男，河北辛集人，研究员．
