Separating oil spill from look-alikes in a SAR image using Artificial Intelligence Techniques

Nearly 84% of global energy requirements are met by fossil fuels, a dominance expected to persist in the coming decades. Untapped oil reserves have been discovered in remote polar regions, and several countries are expanding their maritime shipping routes to explore and transport them. This significantly heightens the risk of maritime oil spills. State-of-the-art remote sensing technologies are used to detect and mitigate oil spill disasters in a timely manner. Remote sensing radars, specifically Synthetic Aperture Radar (SAR), suffer from the drawback of falsely identifying ‘look-alikes’ (e.g., biogenic films and upwelling zones) as true oil spills. To address the challenge of radar backscatter 'look-alikes', we implement automated feature-space learning and deep learning classification schemes. Our optimized Support Vector Machine (SVM) and custom Multilayer Perceptron (MLP) classifiers successfully discriminate real spills from look-alikes using extracted numerical radar features (e.g., intensity, texture, and edge metrics), achieving a peak classification accuracy of 83%. Feature importance and explainability analyses (via SHAP) underscore that boundary gradient descriptors are vital for feature-space performance, preventing accuracy from dropping to 70%. Furthermore, evaluating automated visual representations directly from raw images via a pretrained ResNet-18 transfer learning framework yields a peak validation accuracy of 95.45%. These integrated frameworks offer a robust, physics-informed, and data-driven approach to operational marine pollution monitoring.

Authors

Institutions

Publication Details

Published
2026-11-20
DOI
https://doi.org/10.22541/essoar.15012545/v1
Primary Topic
Oil Spill Detection and Mitigation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Separating oil spill from look-alikes in a SAR image using Artificial Intelligence Techniques

Mihir Kumar Dash, Nairita Pal, Soumya Dutta, Binay Bhattacharya et al.
2 links
Oil Spill Detection and Mitigation
preprint

Separating oil spill from look-alikes in a SAR image using Artificial Intelligence Techniques

Mihir Kumar Dash, Nairita Pal, Soumya Dutta, Binay Bhattacharya, Triparna Sanyal
preprint en

Abstract

Nearly 84% of global energy requirements are met by fossil fuels, a dominance expected to persist in the coming decades. Untapped oil reserves have been discovered in remote polar regions, and several countries are expanding their maritime shipping routes to explore and transport them. This significantly heightens the risk of maritime oil spills. State-of-the-art remote sensing technologies are used to detect and mitigate oil spill disasters in a timely manner. Remote sensing radars, specifically Synthetic Aperture Radar (SAR), suffer from the drawback of falsely identifying ‘look-alikes’ (e.g., biogenic films and upwelling zones) as true oil spills. To address the challenge of radar backscatter 'look-alikes', we implement automated feature-space learning and deep learning classification schemes. Our optimized Support Vector Machine (SVM) and custom Multilayer Perceptron (MLP) classifiers successfully discriminate real spills from look-alikes using extracted numerical radar features (e.g., intensity, texture, and edge metrics), achieving a peak classification accuracy of 83%. Feature importance and explainability analyses (via SHAP) underscore that boundary gradient descriptors are vital for feature-space performance, preventing accuracy from dropping to 70%. Furthermore, evaluating automated visual representations directly from raw images via a pretrained ResNet-18 transfer learning framework yields a peak validation accuracy of 95.45%. These integrated frameworks offer a robust, physics-informed, and data-driven approach to operational marine pollution monitoring.

Technion – Israel Institute of Technology (IL), Indian Institute of Technology Kanpur (IN)
Life below water
Oil Spill Detection and Mitigation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon