摘要
针对Kinect摄像头在追踪手部骨骼关节时识别精度低的问题,提出一种用于纠正该跟踪问题的手部关节位置推断方法。通过Kinect获取深度信息并进行预处理,获得人体区域的深度图像;通过阈值分割获得手臂部分的深度图像;基于正向运动学算法获得腕部关节的3D空间位置,提取出手部的深度图像后利用质心公式计算手部关节的空间坐标。实验结果证明,该方法可以在测试者挥动手臂时较为准确地推断出手部关节位置以及腕部关节和肘部关节的几何关系。
Aiming at the problem that the recognition accuracy of the Kinect camera is low when tracking the hand joints,this paper presents a position inferring method of hand joints for correcting the tracking problem.The depth information was obtained by Kinect and preprocessed to get the depth image of human body region;the depth image of the arm part was obtained by threshold segmentation;based on the forward kinematics algorithm,the 3 D spatial position of wrist joint was obtained;after extracting the depth image of hand,the spatial coordinate of hand joint was calculated by the centroid formula.The experimental results show that our algorithm can more accurately infer the position of the hand joints and the geometric relationship between the wrist joints and the elbow joints when the tester swings the arms.
作者
郭小焕
全太锋
潘英杰
马浩
Guo Xiaohuan;Quan Taifeng;Pan Yingjie;Ma Hao(Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Key Laboratory of Signal and Information Processing of Chongqing,Chongqing 400065,China)
出处
《计算机应用与软件》
北大核心
2020年第10期162-166,共5页
Computer Applications and Software
基金
国家自然科学基金项目(61671095)。
关键词
三维空间
阈值分割
深度信息
手部关节
位置推断
3D space
Threshold segmentation
Depth information
Hand joints
Position inferring
