基于3D建模的手势识别在机械臂抓取中的应用  

Research on object grasping by manipulator based on 3D modeling gesture recognition

作　　者：张志芬[1] 吴怀宇[1] 程磊[1] 杜钊君[1] 

机构地区：[1]武汉科技大学信息科学与工程学院,湖北武汉430081

出　　处：《计算机工程与设计》2014年第6期2147-2152,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(60705035;61075087;61203331;61005065);湖北省科技计划自然科学基金重点项目(2010CDA005)

摘　　要：为解决基于视觉的手势识别易受光照、阴影等因素干扰导致识别率及鲁棒性不稳定的问题,提出了一种基于深度图像传感器的手势识别技术。采用肤色建模与深度图像相结合的方法实现复杂背景下手势的定位分割;采用3D建模实现骨骼点的手势识别;将该算法应用于面向移动机械臂平台的人机交互控制系统中,成功地实现了目标物体的抓取。实验结果表明,该方法识别速度快、准确性高且具有良好的鲁棒性。To solve the problem that the accuracy and robustness of vision-based gesture recognition were unstable due to its susceptibiling to the illumination,shadow and other factors,a gesture recognition technology which based on RGB-D was proposed.The method combining skin-modeling with RGB-D is used to achieve gesture location and segmentation,and on the basis of that,the 3D gesture modeling-based skeleton points were used to achieve gesture recognition.At last,the gesture recognition system was successfully used in a humar-robot interaction framework oriented to mobile manipulator platform to accomplish object grasping tasks.Experimental results showed that the algorithm was realized with characteristics of rapidity,accuracy and good robustness.

关 键 词：肤色建模 深度图像 3D建模 手势识别 人机交互 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]