Relative Pose Estimation for Multi-Camera Systems from Affine Correspondences

Guan Banglei, Ji Zhao*, Daniel Barath, Friedrich Fraundorfer

*Korrespondierende/r Autor/-in für diese Arbeit

Publikation: ArbeitspapierPreprint

Abstract

We propose four novel solvers for estimating the relative pose of a multi-camera system from affine correspondences (ACs). A new constraint is derived interpreting the relationship of ACs and the generalized camera model. Using the constraint, it is shown that a minimum of two ACs are enough for recovering the 6DOF relative pose, ie, 3D rotation and translation, of the system. Considering planar camera motion, we propose a minimal solution using a single AC and a solver with two ACs to overcome the degenerate case. Also, we propose a minimal solution using two ACs with known gravity vector, eg, from an IMU. Since the proposed methods require significantly fewer correspondences than state-of-the-art algorithms, they can be efficiently used within RANSAC for outlier removal and initial motion estimation. The solvers are tested both on synthetic data and on real-world scenes from the KITTI benchmark. It is shown that the accuracy of the estimated poses is superior to the state-of-the-art techniques.
Originalspracheenglisch
PublikationsstatusVeröffentlicht - 21 Juli 2020

Publikationsreihe

NamearXiv.org e-Print archive
Herausgeber (Verlag)Cornell University Library

Fingerprint

Untersuchen Sie die Forschungsthemen von „Relative Pose Estimation for Multi-Camera Systems from Affine Correspondences“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren