The article ‘Bayesian Filtering for Homography Estimation’ published in the IEEE Robotics and Automation Letters in October 2023, presents a study on homography estimation using a Bayesian filtering framework with rate gyro and camera measurements. The researchers, Arturo Del Castillo Bernal, Philippe Decoste, and James Richard Forbes, demonstrate that the use of rate gyro measurements allows for a more reliable estimation of homography, especially when occlusions are present. The application of Bayesian filtering also produces an estimate of uncertainty, which is valuable for adaptive filtering methods, post-processing procedures, and safety protocols. The study tests an iterative extended Kalman filter and an interacting multiple model (IMM) filter using both simulated and experimental datasets. The IMM filter showed good consistency properties and improved overall performance compared to the state-of-the-art homography nonlinear deterministic observer in the simulations and experiments conducted.
Publication date: 18 Oct 2023
Project Page: DOI: see top of this page
Paper: https://arxiv.org/pdf/2310.10612
