Relative Pose Estimation
Overview
Relative pose estimation determines how one camera, object, or robot is positioned and oriented with respect to another, supporting visual odometry, 3D reconstruction, and cooperative robotics. Classical methods use image correspondences, epipolar geometry, minimal solvers, and RANSAC, while learned matchers and geometry models predict or refine poses; filtering and optimization can combine these estimates with inertial, range, or other sensor data. Methods are evaluated on indoor and outdoor image pairs, driving and visual-odometry sequences, and multi-robot settings, with attention to accuracy, runtime, and robustness to outliers or poor visibility. A central challenge is balancing fast learned predictions with geometric reliability and handling scale, camera calibration, motion, and changes in environment across different sensors and applications.
Latest papers
66 papers8 papers in the last four weeks, up 14% on the four weeks before. 0.0% of all new papers.