A New Statistical Model of Star Speckles for Learning to Detect and Characterize Exoplanets in Direct Imaging Observations

2025-03-21

AI Summary

In direct imaging, exoplanet detection is severely hindered by residual stellar light—particularly stellar speckle noise—which overwhelms faint planetary signals. Method: This paper proposes a multi-scale, physics-driven joint spectral-channel statistical model. It introduces a novel stellar speckle modeling framework that integrates rotational symmetry priors with physically grounded spectral constraints, embedded within a differentiable end-to-end learning architecture for simultaneous planet detection and flux estimation. Contribution/Results: Unlike conventional speckle-subtraction methods, the model markedly reduces reliance on high signal-to-noise ratio (SNR) and high-fidelity data, thereby unlocking the scientific utility of lower-quality observations. Evaluated on real SPHERE/VLT data, it achieves superior precision–recall trade-off balance, demonstrates computational efficiency and robustness, and is scalable to large-volume survey operations.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web search models and ranking
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

The search for exoplanets is an active field in astronomy, with direct imaging as one of the most challenging methods due to faint exoplanet signals buried within stronger residual starlight. Successful detection requires advanced image processing to separate the exoplanet signal from this nuisance component. This paper presents a novel statistical model that captures nuisance fluctuations using a multi-scale approach, leveraging problem symmetries and a joint spectral channel representation grounded in physical principles. Our model integrates into an interpretable, end-to-end learnable framework for simultaneous exoplanet detection and flux estimation. The proposed algorithm is evaluated against the state of the art using datasets from the SPHERE instrument operating at the Very Large Telescope (VLT). It significantly improves the precision-recall trade-off, notably on challenging datasets that are otherwise unusable by astronomers. The proposed approach is computationally efficient, robust to varying data quality, and well suited for large-scale observational surveys.

Problem

• Detect faint exoplanet signals in strong starlight
• Model nuisance fluctuations using multi-scale approach
• Improve precision-recall trade-off in exoplanet detection

Innovation

• Multi-scale statistical model for star speckles
• Interpretable end-to-end learnable detection framework
• Joint spectral channel representation with physics