Detection and Tracking of Near-Earth Objects (NEOs): An Approach with Refined Astrometry and Differential Photometry in Python

Abstract

Copyright ©2025 by Instituto Tecnológico de Aeronáutica.The detection and study of Near-Earth Objects (NEOs) are vital for Solar System mapping and planetary defence. This work introduces a fully automated pipeline to process astronomical FITS images, enabling the identification of moving objects, astrometric calibration, and photometric measurements. The method integrates physical-optical models with metaheuristic optimisation, leveraging a hybrid strategy that couples the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to Bayesian Optimisation with Gaussian Processes (BO-GP). Implemented in Python, the framework dynamically tunes all relevant hyperparameters, aiming to maximise completeness and purity while minimising false alarms and computational overhead. Beyond detection, the pipeline incorporates sub-pixel astrometric refinement, automatic association of tracklets, differential photometry, and statistical validation through ANOVA and stratified Wilcoxon tests. Validation was performed with observational sequences from the Pan-STARRS2 telescope, including 160 sets for calibration and 100 independent sets for testing, amounting to over 1000 FITS frames. The system reached 86.4% completeness (recall), 90.5% purity (precision), and an overall F1-score of 0.88, which is on par with established tools such as ZTF MOPS and HelioLinC. Moreover, compared with traditional exhaustive searches, the hybrid approach achieved a 24-fold reduction in median calibration time, thus making near real-time operation feasible. The results highlight the pipeline as a robust, efficient, and scalable solution for automated NEO monitoring in space surveillance contexts.

Keywords