Minimal Cases for Computing the Generalized Relative Pose using Affine Correspondences

TL;DR

This paper introduces three new minimal solvers for estimating the relative pose of multi-camera systems from affine correspondences, improving efficiency and accuracy over existing methods, especially in planar motion scenarios.

Contribution

The paper presents novel solvers that require fewer correspondences for relative pose estimation, including solutions for planar motion and known vertical direction, enhancing robustness and efficiency.

Findings

Solvers outperform state-of-the-art in accuracy on synthetic and real data.

Fewer correspondences needed, enabling faster RANSAC integration.

Effective in planar motion and with known vertical direction scenarios.

Abstract

We propose three 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, we demonstrate efficient solvers for two types of motions assumed. Considering that the cameras undergo planar 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 vertical direction, e.g., 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 odometry benchmark. It is shown that the accuracy of the estimated poses is superior to the state-of-the-art techniques.