三维人手关节点的提取研究

Research of Extracting 3D Joints from Human Hands

近年来,以计算机为载体的人机交互愈发成为人们日常生活中的一个重要组成部分,人手作为仅次于语言的交流方式,有着不可或缺的作用,在人机交互中作为主要媒介有着重要的研究价值;针对在三维手势的多样性问题,在关节点的提取上基于特征匹配提出了一种新的关节点提取研究方法:跨维匹配-即二维数据与三维数据之间的匹配;在立体视觉手势交互的基础上,算法首先用kinect相机采集带有关节点标记的彩色图像和三维点云,然后通过特征点提取,获得手势点云指尖部分的坐标,最后进行匹配得到手势关节点的三维坐标;实验表明采用该方法来提取人手的关节点是可行的,并且在手势关节点的提取效果上也很精确,对研究手势的多样性有着非常重要的作用。

In recent years,the computer as the carrier of human computer interaction has become an important part in People＇s Daily life,hand as a second language of communication has the indispensable role,as the main medium in the human-computer interaction has important research value.In this paper,a kind of new research methods is proposed based on the joint point of the extraction of feature matching.The Kinect was used to collect the same gesture with joint point mark color images and 3D point cloud based on stereo vision gesture interaction,and through the methods of extracting and matching feature points,obtained gestures screen coordinates of the point cloud plan,and compared with 3D gesture point cloud by one-to-one correspondence to get key points of 3 d coordinate.Experiments show that the method is feasible,and the key points in the gestures of extraction effect is also very accurate and.can be easily applied to the tansform of the gestures.