Automated Ship Detection and Characterization in Sentinel-2 Images: A Comprehensive Approach

Bou-Laouz Moujahid (1) , Vadaine Rodolphe (2) , Hajduch Guillaume (2) , Ronan Fablet (1, 3, 4)

1 IMT Atlantique - MEE - Département Mathematical and Electrical Engineering
2 CLS - Collecte Localisation Satellites [Brest]
3 Lab-STICC_OSE - Equipe Observations Signal & Environnement
4 ODYSSEY - Océan Dynamique Observations Analyse

The automatic detection and characterization of ships in optical remote sensing images is a key challenge for maritime surveillance applications. This paper presents an automated system specifically designed for ship detection in medium-resolution Sentinel-2 images. The proposed approach relies on a deep learning model trained on a dataset comprising over 6000 annotated Sentinel-2 images. It achieves a detection rate of 93%, with an average of 2.1 to 3.9 false alarms per Sentinel-2 image. Besides the detection task, it also addresses the estimation of ship lengths as well as ship headings. It yields a mean error of 15.36m ± 19.57m for ship lengths, and estimates ship headings with an accuracy of 93%. This contribution significanly enhances the performance of ship detection and characterization systems in optical remote sensing imagery.

Mots clés: Deep Neural Network; Sentinel-2; Ship Detection; Ship Characterization; Faster R-CNN; Medium Resolution; Object Detection

Domaines: Informatique [cs]; Physique [physics]; Statistiques [stat]

Bou-Laouz Moujahid, Vadaine Rodolphe, Hajduch Guillaume, Ronan Fablet. Automated Ship Detection and Characterization in Sentinel-2 Images: A Comprehensive Approach. 2023. ⟨hal-04359761⟩