Leveraging Synthetic Data for Star and Satellite Photometry

In the realm of Space Domain Awareness (SDA), precise photometric measurements are essential for applications such as stability analysis, shape recovery, and material studies of satellites. However, current methods that rely on manual data collection and analysis are not scalable to autonomous frameworks, which are increasingly necessary due to the growing congestion in space. This research presents an approach to automate photometric measurements within a network of telescopes operating in non-ideal conditions. Our work focuses on achieving reliable photometry in degraded weather conditions, where traditional methods might fail, leading to false detections and unnecessary follow-up efforts. We utilize the SatSim space scene simulator to generate synthetic data for training and testing photometry algorithms. These algorithms include both traditional aperture photometry and machine learning-based approaches. Our methodology employs dynamic segmentation techniques to optimize the detection of satellites and stars under various adverse conditions. The segmentation methods were evaluated for their robustness in different scenarios, with the Depth-First Search + Interquartile Range (DFS + IQR) approach showing the most promise. Through extensive experimentation, we demonstrate that our approach can achieve a photometric precision of approximately 10sup−1/sup, even in adverse conditions. This represents a significant advancement in the field, as it enables more reliable satellite detection and tracking in real-world, non-photometric environments. Additionally, our ablation studies highlight the importance of balanced datasets in reducing error metrics, particularly for underrepresented satellite classes. This work contributes to the development of more effective autonomous SDA systems, capable of operating efficiently in a wide range of environmental conditions.