Bound and Conquer: Improving Triangulation by Enforcing Consistency

We study the accuracy of triangulation in multi-camera systems with respect to the number of cameras. We show that, under certain conditions, the optimal achievable reconstruction error decays quadratically as more cameras are added to the system. Furthermore, we analyze the error decay-rate of major state-of-the-art algorithms with respect to the number of cameras. To this end, we introduce the notion of consistency for triangulation, and show that consistent reconstruction algorithms achieve the optimal quadratic decay, which is asymptotically faster than some other methods. Finally, we present simulations results supporting our findings. Our simulations have been implemented in MATLAB and the resulting code is available in the supplementary material, which can be found on the Computer Society Digital Library at http://doi.ieeecomputersociety.org/10.1109/TPAMI.2019.2939530.

[1]  Karl C. Walli,et al.  Plenoptic camera range finding , 2013, 2013 IEEE Applied Imagery Pattern Recognition Workshop (AIPR).

[2]  Richard I. Hartley,et al.  Multiple-View Geometry Under the {$L_\infty$}-Norm , 2008, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  David J. Kriegman,et al.  Practical Global Optimization for Multiview Geometry , 2006, International Journal of Computer Vision.

[4]  Michael Felsberg,et al.  Robust Three-View Triangulation Done Fast , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops.

[5]  Michael M. Zavlanos,et al.  Exact bias correction and covariance estimation for stereo vision , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[6]  Bernhard P. Wrobel,et al.  Multiple View Geometry in Computer Vision , 2001 .

[7]  Wilfried Philips,et al.  Point Triangulation through Polyhedron Collapse Using the l∞ Norm , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[8]  Vivek K. Goyal,et al.  Quantized Overcomplete Expansions in IRN: Analysis, Synthesis, and Algorithms , 1998, IEEE Trans. Inf. Theory.

[9]  Alexandru Tupan,et al.  Triangulation , 1997, Comput. Vis. Image Underst..

[10]  Klas Nordberg.  Efficient Three-view Triangulation Based on 3D Optimization , 2008, BMVC.

[11]  Zoran Cvetkovic.  Source coding with quantized redundant expansions: accuracy and reconstruction , 1999, Proceedings DCC'99 Data Compression Conference (Cat. No. PR00096).

[12]  Richard I. Hartley,et al.  In Defense of the Eight-Point Algorithm , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[13]  Richard Pollack,et al.  Upper bounds for configurations and polytopes inRd , 1986, Discret. Comput. Geom..

[14]  Antonio Ortega,et al.  EFFICIENT QUANTIZATION FOR OVERCOMPLETE EXPANSIONS IN , 2000 .

[15]  James Arvo,et al.  Fast Random Rotation matrices , 1992, Graphics Gems III.

[16]  Antonio Ortega,et al.  Efficient quantization for overcomplete expansions in RN , 2003, IEEE Trans. Inf. Theory.

[17]  Frederik Schaffalitzky,et al.  How hard is 3-view triangulation really? , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[18]  Kenichi Kanatani,et al.  Triangulation from Two Views Revisited: Hartley-Sturm vs. Optimal Correction , 2008, BMVC.

[19]  Yongduek Seo,et al.  Verifying Global Minima for L2 Minimization Problems in Multiple View Geometry , 2012, International Journal of Computer Vision.

[20]  Hans D. Mittelmann,et al.  An independent benchmarking of SDP and SOCP solvers , 2003, Math. Program..

[21]  Martin Vetterli,et al.  On the accuracy of point localisation in a circular camera-array , 2015, 2015 IEEE International Conference on Image Processing (ICIP).

[22]  Steven D. Blostein,et al.  Error Analysis in Stereo Determination of 3-D Point Positions , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[23]  R. Hartley.  Triangulation, Computer Vision and Image Understanding , 1997 .

[24]  Kenichi Kanatani,et al.  Optimal Two-View Planar Scene Triangulation , 2010, ACCV.

[25]  Martin Vetterli,et al.  Shape: Linear-time camera pose estimation with quadratic error-decay , 2016, 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[26]  Jake K. Aggarwal,et al.  Stochastic Analysis of Stereo Quantization Error , 1990, IEEE Trans. Pattern Anal. Mach. Intell..

[27]  Roman P. Pflugfelder,et al.  Localizing non-overlapping surveillance cameras under the L-Infinity norm , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[28]  Richard I. Hartley,et al.  Optimal Algorithms in Multiview Geometry , 2007, ACCV.

[29]  Sundeep Rangan,et al.  Recursive consistent estimation with bounded noise , 2001, IEEE Trans. Inf. Theory.

[30]  David J. Kriegman,et al.  Practical Global Optimization for Multiview Geometry , 2006, ECCV.

[31]  Hongdong Li,et al.  A practical algorithm for L triangulation with outliers , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[32]  Sang Wook Lee,et al.  Outlier Removal by Convex Optimization for L-Infinity Approaches , 2009, PSIVT.

[33]  Alexander M. Powell,et al.  Error Bounds for Consistent Reconstruction: Random Polytopes and Coverage Processes , 2014, Found. Comput. Math..

[34]  Martin Byröd,et al.  Fast Optimal Three View Triangulation , 2007, ACCV.