Tanker Hull Marking and Name-Change Detection
Very-high-resolution optical imagery at sub-metre GSD can resolve painted characters on tanker transoms and hull sides, making name and IMO-number changes detectable across multi-date image stacks. This page explains which sensors do the work, what they can and cannot read, and how change detection is applied.
Sensors
- Maxar WorldView-3: 0.31 m panchromatic GSD (native; released to 0.31 m for government customers, 0.31–0.50 m commercially depending on licence). Eight SWIR bands and eight VNIR multispectral bands. Revisit approximately 1 day at mid-latitudes with off-nadir tasking. The operative sensor for resolving individual painted characters at typical transom heights of 0.4–1.0 m.
- Airbus Pleiades Neo: 0.30 m panchromatic GSD, four-satellite constellation achieving revisit under 24 hours globally with aggressive off-nadir scheduling. Radiometric quality and dynamic range suit high-contrast hull markings. Stereo and tri-stereo modes available for 3-D hull geometry if needed.
- Planet SkySat: 0.50 m panchromatic GSD (approximately 0.72 m native resampled to 0.50 m). Constellation of around 21 satellites. Useful for confirming vessel presence and broad livery colour changes; character-level legibility is marginal at this resolution and depends on character size and image quality. Best used to cue WorldView-3 or Pleiades Neo tasking rather than as the primary reading sensor.
- SPOT-7: 1.5 m panchromatic GSD. Insufficient to resolve individual painted characters on a standard tanker transom, but useful for monitoring anchorage patterns and confirming vessel identity through hull outline and proportions when higher-resolution tasking is pending.
What a painted character actually looks like from 617 km
WorldView-3 orbits at roughly 617 km. At 0.31 m GSD, a capital letter 400 mm tall occupies about 1.3 pixels in height. That is tight. A letter 600 mm tall, common on VLCC transoms, occupies roughly 2 pixels. Legibility at that scale depends on contrast between the painted character and the hull background, image quality (MTF, signal-to-noise ratio), and collection geometry. Nadir or near-nadir passes give the clearest transom view; oblique angles beyond about 25 degrees off-nadir begin to foreshorten the transom plane and reduce effective resolution.
In practice, analysts do not rely on pixel-by-pixel character recognition alone. The task is change detection: does the string of characters on the transom in the current image match the string in the archive image of the same vessel? A complete repaint, a partial overpaint, or a character addition or deletion all produce a detectable radiometric difference even when individual characters are ambiguous. The comparison is vessel-specific, run against a registered historical image of the same hull at the same location class (anchorage, port berth, STS zone).
The evasion tradecraft and its optical signatures
Sanctions-evading operators typically alter one or more of: the vessel name on the transom and bow, the IMO number (a seven-digit Lloyd's Register identifier that is, in principle, permanent), the port of registry lettering, and the hull livery colour or funnel marking. Each leaves a different optical signature.
Fresh paint over existing characters produces a slightly different spectral reflectance from the surrounding aged hull coating, particularly in the near-infrared bands available on WorldView-3. The overpaint boundary is often visible as a rectangular patch of uniform, high-reflectance paint against a weathered background. A complete hull repaint in a new colour is the most obvious change and detectable even at SPOT-7 resolution. Partial character substitution, where only one or two letters are changed, is the hardest case: the altered characters may be fresher and slightly brighter, but the difference is small and can be masked by viewing angle, sea spray, or rust.
IMO numbers are the most valuable target. They are physically smaller than vessel names, typically 200–300 mm character height on many tankers, which places them right at the detection limit of 0.31 m imagery. Confirmation often requires multiple independent collections under good atmospheric conditions.
Building the comparison: archive depth and image registration
Change detection across a multi-date image stack requires that images of the same vessel be geometrically registered to one another. For a vessel at anchor this is tractable: the hull is a rigid object and its transom plane can be orthorectified using the vessel's known dimensions from Lloyd's or IHS Fairplay records. Registration error below 0.5 pixels is achievable with careful tie-point selection on stable hull features such as portholes, hawse pipes, or deck fittings.
Archive depth matters enormously. Maxar's WorldView archive extends back to WorldView-1 (2007) and WorldView-2 (2009), giving over fifteen years of potential baseline imagery. Pleiades archive runs from 2012. If a vessel's pre-sanction identity is documented in that archive, a baseline image exists. The practical constraint is tasking: archive coverage of any specific anchorage is sparse unless that location was previously tasked. Purpose-built monitoring programmes task high-interest anchorages on a regular cadence, building the comparison stack deliberately rather than relying on opportunistic archive hits.
Cloud cover is the operational enemy. Optical sensors cannot see through cloud. In high-humidity anchorage regions such as the Strait of Malacca or the Gulf of Guinea, cloud-free windows may be separated by days. This is an honest limit of the method: a vessel can complete a name change and depart during a cloud-obscured gap. SAR-based vessel detection (covered on a sibling page) provides continuity of presence monitoring between optical windows, but SAR at current commercial resolutions cannot resolve painted characters.
Spectral bands and the fresh-paint problem
WorldView-3's eight SWIR bands (1195–2365 nm) are not primarily used for character reading, but they do discriminate paint types and ages. Titanium dioxide-based white paints, common for vessel markings, have a distinctive SWIR reflectance profile that differs from aged grey or black hull coatings. A fresh white overpaint on a dark transom produces a detectable SWIR anomaly even when the panchromatic image is ambiguous about the exact characters present. This is not character recognition; it is evidence of recent painting activity, which raises the investigative priority of a vessel for closer examination.
Multispectral band ratios can also distinguish hull livery colour changes. A tanker that was previously black-hulled and is now grey, or previously carried a red waterline band and now does not, will show a clear spectral change between archive and current imagery. Livery changes alone are not proof of identity fraud but are a screening signal that prompts deeper investigation.
Honest limits: what this method cannot do
Sub-metre optical imagery cannot read IMO numbers painted at less than roughly 250 mm character height with confidence, even under ideal conditions. It cannot see through cloud, haze, or sea smoke. It cannot image a vessel underway at speed with the same transom clarity as one at anchor, because motion blur and wake spray degrade the transom view. It cannot confirm identity on its own: imagery evidence of a marking change must be correlated with AIS history, port state records, and vessel dimension data to constitute a finding rather than an indicator.
There is also a timing problem. The Lloyds/IHS Fairplay vessel database, which provides the ground truth for what a vessel's markings should say, is updated periodically but not in real time. A vessel that has legitimately changed its name through proper channels may appear to show a discrepancy until the database catches up. Analysts must cross-check against multiple registry sources before asserting fraud.
Satellize's analytics team applies this multi-source cross-check as standard, drawing on commercial tasking of WorldView-3 and Pleiades Neo alongside archive screening.
From image to finding: the analytic workflow
A typical investigation starts with a vessel of interest identified through AIS anomaly or flag-state screening. Historical imagery of that vessel is retrieved from the Maxar or Airbus archive, filtered for cloud cover and off-nadir angle. A baseline marking profile is extracted: character string, font geometry, paint colour, and approximate character height. A new collection is tasked at a known anchorage or port call.
The new image is registered to the baseline, and a pixel-level difference image is computed on the panchromatic band. Regions of significant change on the transom plane are flagged and reviewed by an analyst. The analyst scores the finding on a confidence scale that accounts for image quality, registration error, and the magnitude of the detected difference. A high-confidence finding, where characters are visibly different and fresh-paint spectral anomalies confirm recent repainting, is reported with annotated image chips and metadata. A low-confidence finding is flagged for re-tasking under better conditions rather than reported as a conclusion.
Typical figures
| Best panchromatic GSD | 0.30 m (Pleiades Neo), 0.31 m (WorldView-3) |
| Minimum legible character height (indicative) | Approximately 400–600 mm under good conditions; 250 mm is at the detection limit and requires multiple collections |
| Revisit cadence (tasked) | Under 24 hours at most latitudes with WorldView-3 or Pleiades Neo off-nadir scheduling; cloud-free windows may extend this to several days in tropical regions |
| Spectral bands used | Panchromatic (450–800 nm) for character reading; VNIR and SWIR (WorldView-3, 8 SWIR bands to 2365 nm) for paint-age and livery-change discrimination |
| Archive depth | Maxar archive from 2007 (WorldView-1); Airbus Pleiades from 2012; SPOT-7 from 2014 |
| Image registration accuracy | Sub-0.5 pixel achievable on rigid hull features with careful tie-point selection |
| Cloud limitation | Optical only; zero penetration of cloud or thick haze. SAR monitoring required for presence continuity between optical windows |
| Delivery latency | Typically 4–24 hours from satellite downlink to analyst-reviewed image chip, depending on ground station and processing pipeline |
| Delivery formats | GeoTIFF image chips, annotated PDF finding report, structured JSON metadata for integration with vessel-tracking databases |
Analytics Satellize can run
| Baseline marking profile | Archive image retrieval, panchromatic orthorectification, manual and semi-automated character extraction | Structured record of vessel name, IMO number string, port of registry text, hull livery colour, and character geometry, stored as a vessel-specific reference file |
| Transom change-detection alert | Multi-date pixel-level difference image on co-registered panchromatic bands, thresholded on the transom region of interest | Automated alert with confidence score, flagging vessels where transom radiometry differs significantly from baseline |
| Fresh-paint spectral anomaly flag | SWIR band ratio analysis (WorldView-3 bands 6–8) comparing current image to baseline reflectance profile on hull surfaces | Per-vessel flag indicating probable recent repainting, delivered as an attribute in the vessel-tracking feed |
| Hull livery change report | Multispectral classification of hull colour zones across multi-date image stack; change in dominant hull colour class triggers report | Annotated image pair (before/after) with colour classification overlay, delivered as PDF and GeoTIFF |
| Analyst-reviewed identity-change finding | Human-in-the-loop review of automated flags, cross-referenced against Lloyd's Register and IHS Fairplay vessel records | Formal finding report with confidence rating (high/medium/low), annotated image chips, and recommended follow-on actions |
| Anchorage monitoring stack | Regular cadence tasking of high-interest anchorages (e.g. Kalamata, Larak Island, Socotra approaches) to build comparison stacks for vessels of interest | Time-series image archive with vessel-level change log, updated on each cloud-free collection |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.