·120 智能系统学报 第7卷 1)提出算法的正确性和有效性得到了验证.对 [11]SAND P,TELLER S.Particle video:long-range motion 于相似算法，需要至少3个点对来进行运动估计，但 estimation using point trajectories[J].International Jour- 是算法的结果很大程度上依赖所取得的3个点对， nal of Computer Vision,2008,80(1):72-91 因而是很敏感、不稳定的。 [12]MATEI B,MEER P.A general method for errors-in-varia- bles problems in computer vision C]//Proc IEEE Conf 2)当有足够多的数据时，所提出的算法是稳定 Computer Vision and Pattern Recognition.Hilton Head, 的，且具有鲁棒性，它可以消除离群值.最后，提出的 USA,2000:18-25. 算法是很简洁的.时间的消耗取决于点对的数量，通 [13]FIROOZFAM P,NEGAHDARIPOUR S.Theoretical accu- 用的算法只需要0.08，近似算法需要大约0.1s. racy analysis of nocular vision systems for scene reconstruc- 参考文献： tion,motion estimation,and positioning problems in com- puter vision[C]//Proc 2nd Int Symposium on 3D Data [1]PAPADIMITRIOU T,STRINTZIS M G,ROUMELIOTIS Processing,Visualization,and Transmission.Washington, M.Robust estimation of rigid body 3-D motion parameters DC,USA,2004:88-895. based on point correspondences[J].IEEE Trans on Cir- [14]SHEIKH Y,HAKEEM A,SHAH M.On the direct esti- cuits and Systems for Video Technology,2000,10(4): mation of the fundamental matrix[C]//Proceedings of the 541-549. IEEE Conference on Computer Vision and Pattem Recogni- [2]WENG J,HUANG T S,AHUJA N.Motion and structure tion.Minnesota,USA,2007:1-7. from two perspective views:algorithms,error analysis,and [15]CHEN P.Why not use the Levenberg-Marquardt method error estimation[J].IEEE Trans on Pattern Analysis and for fundamental matrix estimation[J].IET Computer Vi- Machine Intelligence,1989,11(5):451-476. sion,2010,4(4):286-294. [3]HARTLEY R.In defense of the 8-point algorithm [C]/ [16]WU HH P,CHANG S H.Fundamental matrix of planar IEEE International Conference on Computer Vision.Cam catadioptric stereo systems[J].IET Computer Vision, bridge,USA,1995:1064-1070. 2010,4(2):85-104. [4]EBRAHIMNEZHAD H,ROBUST H G.Motion from space [17]SZELISKI R.Computer vision:algorithms and applications curves and 3D reconstruction from multiviews using perpen- [M].Berlin:Springer-Verlag,2010:343-374. dicular double stereo rigs[J].Image and Vision Compu- 作者简介： ting,2008,26(10):1397-1420. 何雨兰，女，1988年生，硕士研究 [5]MONTIEL J MM,TARDOHS J D,MONTANO L.Struc- 生，主要研究方向为3-D重建算法。 ture and motion from straight line segments [J].Pattem Rec0 gnition,2000,33(8):1295-1307. [6]STEINBACH E G,EISERT P,GIROD B.Model-based 3-D shape and motion estimation using sliding textures[C] Proc Vision Modeling and Visualization Conference.Berlin, Germany,2001:375-382. 顾人舒，女，1989年生，硕士研究 [7]NEUMANN J,FERMULLER C,ALOIMONOS Y.Polydioptric 生，主要研究方向为3-D重建算法， camera design and 3D motion estimation[C]//Proc IEEE Conf Computer Vision and Pattern Recognition.Madison,USA, 2003:294-301. [8]ROACH J W,AGGARWAL J K.Determining the move- ment of objects from a sequence of images[J].IEEE Trans on Pattern Anal and Machine Intell,1980,7(4):554- 袁杰，男，1975年生，副教授，主要 562. 研究方向为3D重建算法、嵌入式系 [9]OLIENSISA J.Multi-frame structure from motion algorithm 统 under perspective projection[J].International Joumal of Computer Vision,1999,34(2):163-192. [10]付永庆，宋宝森，吴建芳.边缘分类SFT算法[J].哈尔 滨工程大学学报，2010,31(5)：163171. FU Yongqing,SONG Baosen,WU Jianfang.Edge classifi- cation of the SIFT alogorithm[J].Journal of Harbin Engi- neering University,2010,31(5):163-171