估计多视点摄像机姿态的两步法.pdfVIP

16 10 Vol. 16 No. 10 2008 10 Opt ics and Precision Engineering Oct. 2008 1004 24X( 2008) 101 8206 1, 2 3 1, 2 1 向长波 , 谢 丹, 刘太辉 , 宋建中 ( 1. 中国科学院长春光学精密机械与物理研究所, 吉林 长春 130033 ; 2. 中国科学院研究生院, 北 10003 ; 3. 湖南师范大学 工学院, 湖南长沙4 10081) : , , , , , 1 pixel , 0. 5, 0. 1 cm, 1 pixel : 摄像机; 姿态估计; 两步法; 三维重建; 三维测量 :T P3 1 : A A twostep algorithm for estimating postures of cameras located in different points of view 1, 2 3 1, 2 1 XIANG Changbo , XIE Dan , LIU Taihui , SON G Jianzhong ( 1. Changchun I nstitute of Op tics , Fine M echanics and Phy sics , Chinese A cademy of S ciences, Changchun 130033, China; 2. Graduate Uni ersity of Chinese A cademy of Sciences , Beij ing 10003 , China; 3. College of E ng ineering , H unan N orm al Uni ersity , Changsha 410081, China) Abstract: A tw ostep algorithm to estimate the postures of cameras located in different points of view w as proposed. T he local 3D model of an object is reconstructed w ith matching points in tw o images. Then, by taking the local 3D models as the center of w orld coordinate system, the posture of every camera is estimated according to the correspondences betw een the 3D feature points and the image fea ture points of a object. The experimental results of synthetic data and real images show that the aver age orientation error, average location error and the reprojection model error of the camera are less than 0. 5, 0. 1cm and 1 pixel, respectively, under the condition of the position error for feature point in less than 1. 0 pixel. Key words: camera; pose estimation; tw ostep method; 3D