TRPLP – Trifocal Relative Pose From Lines at Points

Ricardo Fabbri (1, 2) , Timothy Duff (3) , Hongyi Fan (4) , Margaret Regan (5) , David da Costa de Pinho (6) , Elias Tsigaridas (7) , Charles Wampler (5) , Jonathan Hauenstein (5) , Peter Giblin (8) , Benjamin Kimia (9) , Anton Leykin (2) , Tomas Pajdla (10)

1 UERJ - Universidade do Estado do Rio de Janeiro [Brasil] = Rio de Janeiro State University [Brazil] = Université d'État de Rio de Janeiro [Brésil]
2 Georgia Institute of Technology [Atlanta]
3 School of Mathematics - Georgia Institute of Technology
4 Brown University
5 UND - University of Notre Dame [Indiana]
6 UENF - Universidade Estadual do Norte Fluminense Darcy Ribeiro
7 OURAGAN - OUtils de Résolution Algébriques pour la Géométrie et ses ApplicatioNs
8 University of Liverpool
9 Brown University - Department of Computer Science
10 CIIRC - Czech Institute of Informatics, Robotics and Cybernetics [Prague]

We present a method for solving two minimal problems for relative camera pose estimation from three views, which are based on three view correspondences of (i) three points and one line and (ii) three points and two lines through two of the points. These problems are too difficult to be efficiently solved by the state of the art Grobner basis methods. Our method is based on a new efficient homotopy continuation (HC) solver, which dramatically speeds up previous HC solving by specializing HC methods to generic cases of our problems. We show in simulated experiments that our solvers are numerically robust and stable under image noise. We show in real experiment that (i) SIFT features provide good enough point-and-line correspondences for three-view reconstruction and (ii) that we can solve difficult cases with too few or too noisy tentative matches where the state of the art structure from motion initialization fails.

Mots clés: Pose estimation; Cameras; Three-dimensional displays; Pipelines; Geometry; Computer vision; Feature extraction; Image matching; Image reconstruction; SIFT features; Point-and-line correspondences; Grobner basis methods; HC methods; Generic cases; Simulated experiments; Three-view reconstruction; TRPLP; Minimal problems; Relative camera; View correspondences; Efficient homotopy continuation solver; Relative camera pose estimation

Domaines: Calcul formel [cs.SC]; Vision par ordinateur et reconnaissance de formes [cs.CV]

Ricardo Fabbri, Timothy Duff, Hongyi Fan, Margaret Regan, David da Costa de Pinho, et al.. TRPLP – Trifocal Relative Pose From Lines at Points. CVPR 2020 - IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun 2020, Seattle / Virtual, United States. pp.12070-12080, ⟨10.1109/CVPR42600.2020.01209⟩. ⟨hal-03105216⟩