← Back to all news
arXiv AI

Onboard Marine Anomaly Detection on $\Phi$sat-2: From Simulation-Based Development to In-Orbit Demonstration

The paper reports on the development, integration, and in-orbit testing of a lightweight marine anomaly detection pipeline for ESA’s Φsat‑2 mission. The system combines sea segmentation, self‑supervised feature encoding, generic anomaly detection, and optional anomaly characterization, and was initially trained on simulated imagery before being retrained on real Level‑1 data to address a mismatch between simulation and in‑orbit observations. The experience underscores the value of simulation for pre‑flight risk reduction, the necessity of robust annotation strategies, and the importance of sensor‑aware design for reliable scientific validation.

Related stories

arXiv AI
Oct 5

On-Board Anomaly Detection for Efficient Marine Environmental Monitoring

The paper presents a marine event detection pipeline for Earth observation satellites that uses a self‑supervised neural network encoder to compress images into a latent space, followed by a machine learning anomaly detection model to spot deviations from normal sea patterns. It compares this approach with traditional algorithms such as Isolation Forest, One‑Class Support Vector Machine, and Local Outlier Factors, demonstrating superior performance. The lightweight, resource‑efficient design is tailored for satellites with limited computational resources and has been integrated into missions like ESA’s Phisat‑2 and Microsoft/Thales Alenia Space IMAGIN‑e, enabling timely alerts and efficient data reduction.

By Thomas Goudemant, Clotilde Szywala, Benjamin Francesconi, Michelle Aubrun, Yves Bobichon, Marjorie Bellizzi, Adrien Girard
diffusion
More like this →
arXiv Computer Vision
Sep 14

AquaCubeAI-Powered Monitoring Turbidity on-board {\Phi}sat-2

arXiv:2609.12744v1 Announce Type: new Abstract: Timely monitoring of coastal water quality is critical for environmental protection, yet conventional satellite workflows rely on downlink and ground p...

By Pietro Di Stasio, Francesca Razzano, Elisa Liparulo, Gabriele Meoni, Nicolas Long\'ep\'e, Deodato Tapete, Paolo Gamba, Gilda Schirinzi, Silvia Liberata Ullo
More like this →
arXiv AI
Oct 9

AI-Based On-Board Maritime Object Detection for Earth Observation Payload Data Reduction on Versal Embedded Hardware

The paper presents an AI-based system for onboard maritime object detection on Earth‑observation satellites, aiming to reduce data volume by selectively downlinking relevant vessel information. A YOLOX‑S detector is trained on a dataset of 68 Maxar scenes with 43 vessel classes, then quantized and deployed on a Versal VC1902 DPU with minimal loss in accuracy and a few‑second processing time per scene. Various downlink modes—metadata only, image crops, tiles, or full scenes—are evaluated, showing that tiles retain 98% of vessels while reducing scene volume to 29%, and crops retain 83% with only 3% of the volume.

By Thomas Goudemant, Aur\'elien Bobey, Omar Hlimi, Marjorie Bellizzi
computer-visionefficiency
More like this →
arXiv Machine Learning
Sep 23

Can We Predict Anomaly Detection Performance from Embedding-Space Geometry?

The paper investigates whether the performance of anomaly detection systems can be predicted without labeled anomalies. For kNN-based detectors, it derives a lower bound on AUC that links detection performance to the separation and variance of inlier and outlier scores, and uses this to analyze how density variation, intrinsic dimensionality, and domain mismatch affect score variability. The authors introduce pseudo‑anomaly probes that provide a reference for estimating relative score separation, and demonstrate through experiments on DCASE benchmarks that these probes enable anomaly‑free model selection to outperform conventional development‑set selection, especially under domain shift.

By Kevin Wilkinghoff, Zheng-Hua Tan
ragbenchmarks
More like this →
arXiv Machine Learning
Aug 11

From Benchmark Performance to Tool Deployment: Human-in-the-Loop Anomaly Detection

arXiv:2608. 07770v1 Announce Type: new Abstract: Automated anomaly detection methods often report strong performance on curated academic benchmarks, but their behavior under real-world industrial conditions is less clear.

By Mike Szklarzewski, CJ George, Gavin Smithson, Christopher Stokes, Dakota Fulp, William M. Jones, Benjamin Wynn, Alexander Ur, Agit Yesiloz, Clint Kallenbach, Mark Swartz, Nathan DeBardeleben, Sharmistha Chakrabarti
benchmarksdata-engineering
More like this →