摘要
在基于计算机视觉技术的非接触式人机交互系统中,为了快速推断使用者指示的目标位置,提出一种无需显式求解摄像机参数的指示位置判别方法。利用目标平面上的已知点及其对应的摄像机成像点,求解目标平面和摄像机像平面间的单应矩阵,将目标平面上相交于指示点的两条直线与像平面上的对应直线联系起来,从而通过检测图像中的特定直线推断指示点的位置。实验表明,在2.5 m的距离上,由单应矩阵映射造成的点对应平均误差为0.73 mm;同时,系统对指示点的估计误差从屏幕中央至边缘上升了2.6%。
To get a rapid estimation of the user′s indication position in the computer vision based non-contact human-computer interaction(HCI) system,an indication position estimation method without explicitly calculating the camera parameters was proposed.The homography matrix was calculated using the known points on the target plane and their corresponding points imaged by the cameras.By the homography relationship,the two lines which intersect at the indication point on the target plane were related with their corresponding lines on the image plane,thus the position of the indication point is estimated by detecting specific lines from images.Experimental results show that the average error of the homographic mapping is 0.73 mm at the distance of 2.5 m.Meanwhile,the amplitude of errors rises by 2.6% from the center to the fringe of the screen.
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2011年第1期110-114,共5页
Journal of Optoelectronics·Laser
基金
国家自然科学基金重点资助项目(60832003)
国家自然科学基金资助项目(60772124)
上海大学研究生创新基金资助项目(SHUCX091022)
关键词
计算机视觉
人机交互
单应矩阵
摄像机模型
computer vision
human-computer interaction
homography matrix
camera model