# Improving Astrometric Precision with MLP-Driven Super-Resolution of Star Maps

## TL;DR

This paper introduces a machine learning method to improve the accuracy of star map simulations for space navigation systems.

## Contribution

A novel MLP-based super-resolution method is proposed for correcting star centroid positioning errors in dynamic star simulators.

## Key findings

- The method reduces star centroid positioning errors by 22.9% on average.
- Inter-star angular distance errors are reduced by 37.5% on average.
- The approach outperforms traditional methods in astrometric precision.

## Abstract

Aiming at the star centroid positioning error in dynamic star simulators, a super-resolution star map correction method is proposed based on a multi-layer perceptron (MLP). A complete technical chain of “system calibration–aberration field modeling–network correction” is constructed to establish a data-driven end-to-end framework for unified modeling and compensation of optical aberrations, assembly deviations, and device discreteness. Experimental results show that the proposed method achieves sub-pixel accuracy: the maximum star centroid and inter-star angular distance errors are reduced by 22.9% and 37.5% on average, respectively, which is significantly superior to traditional methods. This work provides a reliable technical approach for high-precision star map display and star sensor ground calibration, with clear engineering application value.