机器视觉中的优化算法研究论述

邹 慧敏
北京电子科技职业学院

摘要

机器视觉是信息科学领域一门前沿的综合交叉学科。优化算法在该领域有着十分重要的应用。本文概括了一 些典型问题如位姿问题、图像匹配问题中的优化算法。位姿问题主要有两种模型，三维正交约束下分别极小化一个二 次目标和另一个非线性目标。针对二次模型，本论文综述：一基于加强版Lagrangian对偶设计一个新的快速算法，二是 进一步设计一种巧妙的分支定界算法，一个隐含的好处是我们可以同时优化两个目标（多目标规划）；图像匹配中我们 着重归纳两个问题，一个是匹配求基本矩阵的优化模型以及设计高效算法，另一个是研究图像识别的二次匹配模型及 相关算法。

关键词

机器视觉；位姿问题；图像匹配问题

参考

[1] C.Lu,G.D.Hager, E. Mjolsness, Fast and globall y convergent pose estimation from video images, IEEE Transactions on Pattern Analysis and Machine Intelligen ce 22 (6) (2000) 610–622 [2] G.Schweighofer, A. Pinz, Robust pose estimatio n from a planar target, IEEE Transactions on Pattern A nalysis and Machine Intelligence 28 (12) (2006)2024–20 30. [3] H. Hmam,J.Kim Optimal non-iterative pose esti mation via convex relaxation Image and Vision Computi ng 28 (2010) 1515–1523 [4] G.Schweighofer,A.Pinz, Globally optimal O(n) s olution to the PnP problem for general camera models, in: British Machine Vision Conference, 2008. [5] 潘绍松,左洪福.基于流形优化法的相机位姿估计 [J].江苏大学学报(自然科学版),32(3)(2011):336-340. [6] 肖永亮,苏显渝,薛俊鹏,刘晓青.基于凸松弛全局 优化算法的视觉测量位姿估计[J].光电子·激光,22(9),(201 1):1384-1389. [7] V. Lepetit, F. Moreno-Noguer, P. Fua, EPnP: an accurate O(n) solution to the PnP problem, Internationa l Journal of Computer Vision (2008). [8] Sameer Agarwal,Manmohan Chandraker, Fredrik Kahl, David Kriegman, and SergeBelongie. Practical gl obal optimization for multiview geometry. In European Conference on Computer Vision, pages 592–605, 2006. [9] R.I. Hartley, F. Kahl, Global Optimization throu gh Rotation Space Search Int J Comput Vis (2009) 82: 64–79. [10] C. Olsson, F. Kahl, and M. Oskarsson. Optim al estimation of perspective camerapose. 18th Internation al Conference on Pattern Recognition, 2:5–8,2006. [11] C. Schellewald. Convex Mathematical Program s for Relational Matching of Object Views. Ph.D. thesis, Department of Mathematics and Computer Science, Uni versity of Mannheim, 2005. [12] C. Schellewald, S. Roth, C. Schnörr, Evaluatio n of a convex relaxation to a quadratic assignment ma tching approach for relational object views, Image and Vision Computing 25,1301–1314,2007. [13] N. Brixius, K. Anstreicher, Solving quadratic a ssignment problems using convex quadratic programing relaxations, Optimization Methods and Software 16 (1–4) (2001):49–68. [14] Y. Xia, An Efficient Continuation Method for Quadratic Assignment Problems,Computers & Operations Research,37(6):1027-1032, 2010. [15] 郑开杰,高玉涛,彭济根.赋权图匹配问题的一种 新的松弛模型[J].自动化学报,36(8)(2010),1200-1203. [16] C. Schellewald, S. Roth, C. Schnörr, Probabili stic Subgraph Matching Based on Convex Relaxation, In Energy Minimization Methods in Computer Vision an d Pattern Recognition Lecture Notes in Computer Scien ce, 2005, Volume 3757/2005, 171-186. [17] Q. Yang, S.-H. Sze, Path Matching and Graph Matching in Biological Networ, Journal of Computation al Biology, 14(1), 2007. [18] H. Longuet-Higgins A computer algorithm for reconstructing a scene from two projections, Nature, 239, (1981), 133-135. [19] R.I. Hartley，In defence of the 8-point algorith m In proceedings of the 5th international conference on computer vision, (1995), 1064-1070. [20] R.I. Hartley In defence of the eight-point algo rithm IEEE Trans. Pattern Analysis and Machine Intellig ence, 19(6), (1997), 580-593