Demonstration of Plate Analysis, an algorithm for precise relative astrometry

Abstract

Astrometry plays a crucial role in understanding the structure, dynamics, and evolution of celestial objects by providing precise measurements of their positions and motions. We propose a new approach to wide-field, relative astrometry, Plate Analysis. Plate Analysis is an innovative algorithm that estimates stellar coordinates and corrects geometric distortion based on precise reference sources, without relying on dedicated calibration fields. It is implemented as a probabilistic framework using Stochastic Variational Inference to efficiently optimize the numerous parameters involved in the model. The methodology was tested through a simplified simulation designed after the Galactic center survey of the JASMINE mission. This simulation, called the JASMINE mini-mock survey, covered three years of observation with 100 satellite orbits, providing a comprehensive dataset for evaluating the performance of Plate Analysis. Although the observation model incorporated more than 30,000 parameters, the parameters were efficiently optimized through Plate Analysis. The results showed that the estimated coordinates closely match the expected stellar motions, with an average positional error of about $70\,\mathrm{\mu as}$. These findings validate the potential of Plate Analysis for precise wide-field astrometric applications, offering significant insights into improving the accuracy of stellar dynamics measurements.