An Estimation Method Robust to Outliers for Multiple Homographies
QI Naixin1, ZHANG Shengxiu1, CAO Lijia2,3, YANG Xiaogang1, SUN Qiao1
1. Department of Control Engineering, Rocket Force University of Engineering, Xi'an 710025, China;
2. Artificial Intelligence Key Laboratory of Sichuan Province, Zigong 643000, China;
3. College of Automation and Information Engineering, Sichuan University of Science & Engineering, Zigong 643000, China
齐乃新, 张胜修, 曹立佳, 杨小冈, 孙巧. 一种对错误匹配点鲁棒的多单应矩阵估计方法[J]. 机器人, 2017, 39(5): 608-619.DOI: 10.13973/j.cnki.robot.2017.0608.
QI Naixin, ZHANG Shengxiu, CAO Lijia, YANG Xiaogang, SUN Qiao. An Estimation Method Robust to Outliers for Multiple Homographies. ROBOT, 2017, 39(5): 608-619. DOI: 10.13973/j.cnki.robot.2017.0608.
摘要针对错误匹配点干扰条件下的多单应矩阵估计问题，提出了一种对错误匹配点鲁棒的多单应矩阵估计初始化方法．该方法基于特征点对的代数误差和结构相似性约束条件，将错误匹配点剔除策略有机地融合到单应矩阵估计的过程中，在不增加计算复杂度的前提下，能够有效地剔除错误匹配点并估计出多单应矩阵的初值．结合AML-COV（approximate maximum likelihood with homography covariance）后端优化算法，本文通过仿真数据实验和真实图像实验从客观性能指标和主观视觉效果方面对算法的性能进行了验证分析．实验结果表明，本文提出的多单应矩阵估计方法能够精确、高效、鲁棒地估计出多单应矩阵的值，较好地解决了错误匹配点干扰条件下的多单应矩阵估计问题．
Abstract：For the multiple homographies estimation problem in the case of outliers, an initialization method of the multiple homographies estimation robust to outliers is proposed. In this method, the outlier rejection is integrated into the multiple homographies estimation based on the algebraic error and the structure similarity constraint of the key-point correspondences. As a result, the outliers can be removed effectively and the initialization value of multiple homographies can be estimated with a negligible computational overhead. Combining the AML-COV (approximate maximum likelihood with homography covariance) algorithm, several experiments based on simulation data and real images demonstrate the performance of the proposed method in subjective visual quality and objective measurement quality. The experimental results show that the proposed method can achieve accurate, efficient, and robust multiple homographies estimation and performs a good solution to the multiple homographies estimation problem in the case of outliers.
[1] 杨磊,李桂菊,王丽荣.面向场景重构的多序列间配准[J].光学精密工程,2015,23(2):557-565. Yang L, Li G J, Wang L R. Registration between multiple sequences for scene reconstruction[J]. Optics and Precision Engineering, 2015, 23(2):557-565.
[2] Salvi J, Matabosch C, Fofi D, et al. A review of recent range image registration methods with accuracy evaluation[J]. Image and Vision Computing, 2007, 25(5):578-596.
[3] Zaragoza J, Chin T J, Brown M S, et al. As-projective-as-possible image stitching with moving DLT[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA:IEEE, 2013:2339-2346.
[4] Rav-Acha A, Pritch Y, Lischinski D, et al. Dynamosaicing:Mosaicing of dynamic scenes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007, 29(10):1789-1801.
[5] 王锋,程敏,陈小平.面向机器人室内建图的RGB-D图像对齐算法[J].机器人,2015,37(2):129-135. Wang F, Cheng M, Chen X P. An RGB-D image alignment algorithm for robotic mapping in indoor environments[J]. Robot, 2015, 37(2):129-135.
[6] Le M H, Trinh H H, Hoang V D, et al. Automated architectural reconstruction using reference planes under convex optimization[J]. International Journal of Control, Automation and Systems, 2016, 14(3):814-826.
[7] Zelnik-Manor L, Irani M. Multiview constraints on homographies[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(2):214-223.
[8] Kim S, Yang D W, Park H W. A disparity-based adaptive multi homography method for moving target detection based on global motion compensation[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2016, 26(8):1407-1420.
[9] Assuja M A, Suwardi I S. 3D coordinate extraction from single 2D indoor image[C]//International Seminar on Intelligent Technology and Its Applications. Piscataway, USA:IEEE, 2015:233-238.
[10] Shashua A, Avidan S. The rank 4 constraint in multiple (≥ q 3) view geometry[C]//4th European Conference on Computer Vision. Berlin, Germany:Springer, 1996:196-206.
[11] Chen P, Suter D. Rank constraints for homographies over two views:Revisiting the rank four constraint[J]. International Journal of Computer Vision, 2009, 81(2):205-225.
[12] Eriksson A, van den Hengel A. Optimization on the manifold of multiple homographies[C]//IEEE 12th International Conference on Computer Vision Workshops. Piscataway, USA:IEEE, 2009:242-249.
[13] Chojnacki W, Szpak Z L, Brooks M J, et al. Multiple homo-graphy estimation with full consistency constraints[C]//Interna-tional Conference on Digital Image Computing:Techniques and Applications. Piscataway, USA:IEEE, 2010:480-485.
[14] Szpak Z L, Chojnacki W, Eriksson A, et al. Sampson distance based joint estimation of multiple homographies with uncalibrated cameras[J]. Computer Vision and Image Understanding, 2014, 125:200-213.
[15] Fischler M A, Bolles R C. Random sample consensus:a paradigm for model fitting with applications to image analysis and automated cartography[J]. Communications of the ACM, 1981, 24(6):381-395.
[16] Hartley R, Zisserman A. Multiple view geometry in computer vision[M]. Cambridge, UK:Cambridge University Press, 2003.
[17] 许可可,朱文球,郭富禄.基于结构相似的RANSAC改进算法[J].计算机工程与应用,2016,52(12):168-171,245. Xu K K, Zhu W Q, Guo F L. Improved RANSAC algorithm based on structural similarity[J]. Computer Engineering and Applications, 2016, 52(12):168-171,245.
[18] Mou W, Wang H, Seet G, et al. Robust homography estimation based on non-linear least squares optimization[C]//IEEE International Conference on Robotics and Biomimetics. Piscataway, USA:IEEE, 2013:372-377.
[19] Ferraz L, Binefa X, Moreno-Noguer F. Very fast solution to the PnP problem with algebraic outlier rejection[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA:IEEE, 2014:501-508.
[20] Cadena C, Carlone L, Carrillo H, et al. Simultaneous localization and mapping:Present, future, and the robust-perception age[EB/OL]. (2016-06-19)[2017-01-03]. https://arxiv.org/pdf/1606.05830v1.pdf@.
[21] Press W H, Teukolsky S A, Vetterling W T, et al. Numerical recipes in C:The art of scientific computing[M]. Cambridge, UK:Cambridge University Press, 1988.
[22] Oxford Visual Geometry Group. Model house datasets[DB/OL]. (2004-09-02)[2017-01-03]. http://www.robots.ox.ac.uk/~vgg/data/data-mview.html.
