Deterministic Riccati Observers for Homography Decomposition

Abstract

An approach to tackle the classical problem of homography decomposition is presented. The key novelty lies in the design of two deterministic Riccati observers tailored specifically the homography decomposition problem, departing from the conventional frame-by-frame solving approach employed in traditional algebraic methods. By leveraging the temporal correlation within image sequences and harnessing the inherent low-pass response of the observers, the resulting estimates are robust and less susceptible to measurement noise. Furthermore, a comprehensive analysis pointing out sufficient conditions of uniform observability ensuring local exponential stability is conducted. Simulation results demonstrate a broad convergence domain and showcase the effective performance of the proposed observers. Additionally, experimental evidence is provided to illustrate the improved performance compared to classical algebraic solutions.