基于 Kinect 深度信息的手指及手部跟踪研究

摘要

This paper presents a scheme,it tracks the fingers and palms timely based on depth information and can be used for controlling the application program interfaces (API)of electronic devices.First it employs Kinect to capture the depth information,and then generates 3D point cloud of hands,as well as filters it to convert to pixel matrix.The next,it uses K-curvature algorithm to obtain fingertips,and then calculates the data such as the length and width of fingers,the orientations of the fingers and palms,etc.Experimental results show that the scheme has stable and efficient recognition and tracking effect,and is unaffected by light and complex background as well.This method can simultaneously track the motion trajectories of two hands with ten fingers and two palms.It can be used to control the API of electronic devices;the user can also input characters by writing freely in the air with the Kinect.%提出一个基于深度信息对手指和手部进行实时跟踪，并可用于控制电子设备 API 的方案。首先使用 Kinect 获取深度信息，然后生成手部的三维点云，进行过滤转换成像素矩阵；再使用 K-curvature 算法获取指尖，进而获取手指长宽、手指朝向和手掌朝向等数据。实验结果证明该方案识别追踪效果稳定且高效，不受光照和复杂背景影响。能够同时跟踪双手共十个手指和两个掌心的动作轨迹，可用于控制电子设备 API，使用者也能用手指在空气中通过 Kinect 输入字符。