基于视觉特征库的交互控制方法  

An Interactive Control Method Based on Vision Feature Library

作　　者：徐增敏[1,2] 关健[2] 蒋英春[1] 段雪峰[1] 

机构地区：[1]桂林电子科技大学数学与计算科学学院,广西桂林541004 [2]武汉大学国家多媒体软件工程技术研究中心,武汉430072

出　　处：《计算机测量与控制》2014年第3期754-757,共4页Computer Measurement &Control

基　　金：国家高技术研究发展计划(863计划)项目(2012AA011005);国家自然科学基金项目(11101100)

摘　　要：针对现有人机交互控制方法在不同应用场景需要采用不同算法,各种样本库训练费时等情况,提出以凸包缺陷面积和来刻画手掌,不需要添加额外样本库,总结出了以物距、手势特征周长、手势特征面积、凸包缺陷三角形面积和等因素为基础的手势特征库,设计出了手势控制的整套算法,并以幻灯片控制为例,使用VS2008和OpenCV完成了交互控制的软件开发;实验结果表明该方法虚拟交互时间在1 ms内,能很好的识别出2种特定的动态手势,解析出多种控制命令,可用于交互系统控制。Aiming at the problems that many application scenes need different interactive control algorithms, and the sample library train- ing cost lots of time, a hand gesture recognition method without extra sample library combine with sum of the convex defects area is pro- posed. The hand gesture library based on object distance, hand gesture perimeter, hand gesture area and sum of the convex defects triangle area is concluded, then a whole hand gesture control algorithms is designed, the Power Point interactive control software is developed by VS2008 and OpenCV as well. Experimental results show that this method which is good at recognizing 2 specific hand gestures, reflect high real--time process in human--machine interaction within l ms, it can be used to interpret many control commands and control interactive sys- tem.

关 键 词：手势识别 视觉特征库 交互控制 凸包缺陷 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]