基于自适应提取和改进CAMSHIFT单目手势跟踪  

Monocular gesture tracking based on adaptive extraction and improved CAMSHIFT

作　　者：黄敦博 林志贤[1] 姚剑敏[1] 郭太良[1] 

机构地区：[1]福州大学物理与信息工程学院,福建福州350002

出　　处：《电视技术》2016年第7期107-112,共6页Video Engineering

基　　金：国家科技部"863"重大专项(2013AA030601);福建省科技重大专项(2014HZ0003-1);福建省资助省属高校专项课题项目(JK2014002)

摘　　要：为了解决复杂背景下手势提取与手势跟踪准确度受影响的问题,提出了一种基于自适应提取和改进CAMSHIFT(Continuously Adaptive Mean Shift)单目手势跟踪算法。该算法通过自适应手部提取方法识别手部完成对跟踪目标的初始化,以运动历史矩阵为掩模提取图像中的感兴趣区域,在该区域内使用CAMSHIFT算法跟踪目标,并通过傅里叶描述子对跟踪目标轮廓进行实时反馈,完成对动态手势的精确跟踪。该方法能够在手部经过肤色区域的情境中实现准确跟踪,与经典CAMSHIFT算法相比,跟踪正确率提高了80%,实现了复杂背景下动态手势的准确跟踪。In order to solve gesture extraction and gesture tracking accuracy affected by the problem of complex background, a mo- nocular gesture tracking algorithm based on the adaptive extraction and improved CAMSHIFT（ Continuously Adaptive Mean Shift） strategy is proposed in this paper. The method is automatically accomplished by the adaptive hand extraction method identifying the hand. The CAMSHIFI＂ algorithm is used to track the target in the area that is extracted by the motion history matrix as a mask ,then Fourier operator calculates the tracked dynamic gestures as real-time feedback to complete the precise tracking of the dynamic ges- ture. The proposed method can track the dynamic gesture accurately under the condition that the motion gestures and facial over- lapped each other. The rate of tracking precision rises 80% which compares with the traditional CAMSHIFT algorithm. The method realizes the precision dynamic tracking under the complex background.

关 键 词：动态手势跟踪 自适应手部提取 运动历史矩阵 改进CAMSHIFT 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]