摘要
基于计算机视觉的人体关节运动跟踪是建立在计算机基础上的可移植、可拓展的智能人机交互模式.通过极值点确定、初始位置确定、粗定位、精定位、带宽确定,获取运动人体关节的中心和带宽,从而识别人体关节.采用改进的Mean Shift算法对人体运动时的关节进行跟踪,应用改进的Kalman滤波器跟踪算法对运动目标的状态进行估计,解决了快速运动或突然变向的目标丢失的问题.最后,对改进前后Kalman滤波器跟踪算法的跟踪效果进行对比,验证了改进Kalman滤波器跟踪算法的优越性.
Human body joint motion tracking based on computer vision is a portable, scalable mode of intelligent human- computer interaction, which is established on the basis of computer. Through the extreme value point determining, the initial position determining, coarse positioning, fine positioning, and band width determining, the center and band width of the movement of human joints is determined, thereby the human joints are recognized. To estimate the position of the next frame of the moving target and solve the problem of the loss of target when rapid movement or a sudden change of human body, the improved Mean Shift algorithm for tracking human body joint and the improved Kalman filter are adopted. Finally, the tracking effect of the Kalman filter tracking algorithm before and after improvement is compared, and the superiority of improved Kalman filter tracking algorithm is verified.
出处
《东华大学学报(自然科学版)》
CAS
CSCD
北大核心
2013年第4期448-454,共7页
Journal of Donghua University(Natural Science)
关键词
计算机视觉
人机交互
人体关节
运动跟踪
识别
computer vision
human-computer interaction
human body joint
motion tracking
recognition