A Self-Calibrating Closed-Loop Framework for Robust Revisit Positioning in Astronomical Imaging

Abstract

Accurate target repositioning is essential in automated astronomical imaging because previously detected sources must be reliably centred during subsequent observations. Although source detection may be accurate during initial acquisition, revisit operations can still produce substantial positioning errors because of coordinate offsets, scale mismatch, cross-axis coupling, calibration drift, and inconsistencies between image and control-coordinate systems. This study aimed to develop and evaluate a self-calibrating closed-loop framework for correcting such revisit-positioning errors. A publicly available Sloan Digital Sky Survey r-band astronomical image was analysed, and 78 compact sources were detected using DAOStarFinder. A controlled revisit fault combining translation, anisotropic scaling, and horizontal–vertical coordinate coupling was then introduced. Three correction strategies were evaluated: affine coordinate calibration, closed-loop preview correction, and online residual learning. The simulated fault produced a mean centring error of 60.89 pixels, with none of the sources positioned within the predefined 10-pixel tolerance. Affine calibration achieved the best performance, reducing the mean error to 1.25 pixels, representing a 97.9% reduction, and positioning all sources within tolerance. Closed-loop preview correction reduced the mean error to 3.13 pixels and also achieved complete tolerance compliance. Online residual learning reduced the mean error to 18.49 pixels but was less effective because the implemented global-bias model could not represent spatially varying distortions. The findings demonstrate that severe revisit-positioning errors can arise from coordinate-calibration and software-integration faults even when source detection is accurate. The proposed framework provides a reproducible basis for combining geometric calibration, preview-based feedback, adaptive correction, and drift monitoring in automated astronomical revisit-imaging systems.