Leveraging satellite imagery to determine Oil spills at sea along with AIS data correlations to identify vessel responsible for the spill.
National Technical Research Organisation (NTRO) · Space Technology · Software
One of the best-value problems in this block — free data, a gripping end-to-end attribution demo, a thin field — but handle the SAR false positives and present the origin as a probability region, because compounding errors and over-precise attribution are where it can fall apart.
Data: AIS Data from marinecadastre.gov/accessais/
What it actually is
Oil spills at sea damage ecosystems and the responsible vessel usually escapes because nobody links the slick to a ship. The ask is a pipeline that detects a spill in satellite radar imagery, traces the slick backward to where and when it started, and then uses ship tracking data to reconstruct which vessels were nearby and score which one is the likely culprit.
What to build
A three-stage pipeline: detect and characterise oil slicks in SAR imagery — Sentinel-1 dark-formation detection with computation of slick area, shape and, where possible, age — then use ocean current and wind data to hindcast the slick's drift backward to an estimated origin point and time and forecast its future spread, and finally reconstruct vessel traffic around that origin window from AIS data, filter out irrelevant traffic, and score suspect vessels on proximity, trajectory alignment with the slick and behavioural anomalies such as an AIS gap, presented on a map interface showing the slick, the drift path and the ranked suspect vessels.
Smallest thing that wins the room
Load a SAR scene with a real oil slick, detect and outline it, hindcast its drift back to an origin point and time, then overlay AIS tracks from that window and show the pipeline ranking three candidate vessels with the top suspect's trajectory hugging the reconstructed origin.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 71% of the 226 · #67 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: defence, intelligence and space bodies drew small fields.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
4/5Underrated — all the data is free and public, the attribution pipeline is a genuinely compelling end-to-end story, and the maritime-forensic domain thins the field, though the SAR false-positive problem and the physical hindcasting are real difficulties you must handle honestly rather than gloss.
Feasibility
3/5Sentinel-1 SAR and AIS data are both free and public, and there are labelled oil-spill SAR datasets, but coupling detection with physical drift hindcasting and AIS correlation is a genuine three-discipline integration, and distinguishing real slicks from the many look-alikes in SAR — algae, low wind, ship wakes — is the hard, error-prone core.
Innovation scope
4/5The drift hindcasting to an origin and the vessel-scoring logic — how you weight proximity against trajectory alignment and AIS anomalies — are genuinely open, and combining the three stages into one attribution pipeline is where the real contribution lies.
Clarity
5/5The description numbers the three tasks precisely — detect and characterise the slick, hindcast drift to origin and forecast spread, attribute to a vessel by reconstructing and scoring AIS traffic — and even names the scoring factors, making the requirement exceptionally clear.
Effort
MassiveSAR detection, physical drift modelling and AIS correlation with suspect scoring are three substantial subsystems in different disciplines that must be chained into one working pipeline.
Demo-ability
EasyA slick detected, drifted back to an origin, and matched to a ranked suspect vessel on a map is a complete, self-explanatory investigative narrative that lands powerfully.
In its favour
- Green flag: Sentinel-1 SAR and AIS data are both entirely free and public, and marinecadastre.gov is linked directly for the AIS format, so the whole pipeline runs on open data
- Green flag: The three-stage attribution story — detect, hindcast, attribute — is one of the most compelling end-to-end narratives in the entire problem set
- Green flag: OpenDrift is a mature open-source Lagrangian drift model, so the physical hindcasting rests on established tooling rather than a hand-rolled model
- Green flag: AIS gaps and course anomalies are genuine behavioural signals, so the suspect scoring has real forensic logic behind it
Against it
- Red flag: SAR oil-slick detection is plagued by look-alikes — biogenic films, low-wind zones, ship wakes — and false positives are the dominant failure mode, so detection precision is where an NTRO judge will press first
- Red flag: Drift hindcasting is inherently uncertain, so the reconstructed origin is a probability region not a point, and presenting it as a precise location overstates what the physics supports
- Red flag: Attributing a spill to a named vessel is a forensic accusation, so the scoring must be presented as ranked suspicion with evidence, not as a determination
- Red flag: Chaining three subsystems means errors compound — a mislocated origin sends the AIS correlation to the wrong ships entirely
What you will be writing
- Sentinel-1 SAR dark-formation detection (U-Net / thresholding)
- Oil spill SAR datasets (Copernicus, published collections)
- OpenDrift / Lagrangian drift hindcasting
- ERA5 wind + ocean current reanalysis
- AIS trajectory reconstruction and anomaly scoring
- Leaflet map with slick, drift path and suspect ranking
- Maritime surveillance
- SAR remote sensing
- Vessel attribution
Prior art to read before you start
SAR oil spill detection · slick drift hindcasting to origin · AIS-based vessel attribution
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.