Deep learning for exoplanet detection and characterization by direct imaging at high contrast

2025-09-24

AI Summary

High-contrast direct imaging of exoplanets faces dual challenges: the requirement for high angular resolution and severe interference from quasi-static speckle noise and instrumental aberrations, hindering reliable detection and precise characterization of faint planetary signals. To address this, we propose a physics-informed, multi-scale statistical model that jointly embeds stellar point-spread function (PSF) modeling and noise generation mechanisms into a differentiable deep learning framework. Our method introduces a multi-frame, signal-to-noise ratio–adaptive fusion strategy and is trained end-to-end on real VLT/SPHERE observational data. It effectively suppresses structured noise while preserving fine spatial detail, thereby enhancing detection sensitivity. Astrometric and photometric accuracies improve by approximately 30% and 25%, respectively, demonstrating robustness and state-of-the-art performance under realistic observing conditions.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Exoplanet imaging is a major challenge in astrophysics due to the need for high angular resolution and high contrast. We present a multi-scale statistical model for the nuisance component corrupting multivariate image series at high contrast. Integrated into a learnable architecture, it leverages the physics of the problem and enables the fusion of multiple observations of the same star in a way that is optimal in terms of detection signal-to-noise ratio. Applied to data from the VLT/SPHERE instrument, the method significantly improves the detection sensitivity and the accuracy of astrometric and photometric estimation.

Problem

• Detecting exoplanets via direct imaging requires high angular resolution and contrast
• Modeling nuisance components corrupting multivariate image series in exoplanet imaging
• Improving detection sensitivity and accuracy of astrometric and photometric estimation

Innovation

• Multi-scale statistical model for nuisance component
• Learnable architecture leveraging physics of problem
• Optimal fusion of multiple star observations