Which subsea 3D method actually fits your inspection campaign?
If you’ve ever tried to scope a subsea inspection campaign that involves dimensional data, you’ll know the first problem isn’t technical, it’s linguistic. “3D scanning” gets used to describe everything from a bulky £300k spread to a GoPro on a stick. The technologies, accuracies, and deliverables behind that label vary enormously, and most of the time the differences only become apparent once the dimensional analysis work begins, or worse, when you come to install the retro-fit part.
We’ve spent ten years ensuring that there is no gap between what we promise and what we deliver, ensuring all answers actually hold up under engineering scrutiny, and retro-fit parts fit first time. The short answer, if you’re in a hurry: the right 3D method for a subsea inspection campaign depends entirely on the engineering , location of the challenge and question you need answered — not the sensor, not the hardware brand, and not the price per day. Laser, stereo, structured light and stills-based photogrammetry each have a legitimate place, but they differ sharply in accuracy, auditability, access constraints and what you actually receive as a deliverable.
This post lays out those differences honestly, maps them against the integrity scenarios where they matter most, and explains why we think the industry needs to talk less about scanning and more about dimensional assurance.
The methods, compared
There are four broad approaches to capturing 3D data subsea. Each has a legitimate place, and none of them is universally better than the others — context decides everything. But the differences in what you get back, and how much you can trust it afterwards, are worth understanding before you write your scope of work.
| Laser Scanning | Structured Light | Stereo Camera Systems | Still-based Photogrammetry | |
|---|---|---|---|---|
| How it works | Measures distance using time-of-flight or phase-shift laser returns, building a point cloud from millions of individual range measurements | Projects a known light pattern onto a surface and calculates geometry from how the pattern deforms — essentially reverse-engineering the shape from distortion | Two fixed cameras at a known baseline capture simultaneous images; software triangulates depth from the parallax between them | Overlapping still photographs from multiple angles are processed to reconstruct full 3D geometry through feature matching and bundle adjustment |
| Typical Subsea Accuracy | Sub-millimetre at nominal range, but degrades with changes in distance and turbidity; refraction through water adds systematic error that requires careful calibration | High accuracy at close range (tenths of a millimetre in ideal conditions), but working envelope is small and sensitive to ambient light and particulates | Geometry-dependent; typically 1–5mm for subsea systems, limited by fixed baseline and image resolution | Sub-millimetre achievable with proper methodology — our Lloyd's Register trial returned 0.03% max error; 50-micron accuracy (1/3000) |
| Auditability | Accuracy depends on pre-job calibration; difficult to verify independently after the fact | Same limitation — calibration-dependent, not independently auditable from the dataset alone | Baseline geometry is fixed at manufacture; post-hoc verification is limited to checking the stereo pair consistency | In-scene scale references (scale boards) are captured in the imagery itself, making accuracy independently verifiable at any point in the future |
| Hardware Dependency | Requires specialist subsea laser unit — typically large, expensive, and not already on the vessel | Requires dedicated projection and capture unit; confined to controlled standoff distances | Requires the manufacturer's stereo rig; no flexibility to swap cameras or adapt to conditions | Camera-agnostic — works with whatever suitable camera is available, including kit already on the vessel or ROV in urgent situation |
| Confined Space Access | Limited by unit size; most systems cannot operate in hawse pipes, caissons, or tight internal structures | Small working envelope suits some confined scenarios, but the projector-camera geometry constrains positioning | Fixed rig geometry means the system either fits or it doesn't — no way to adapt if the space is tighter than expected | A single camera and lighting rig can access spaces where nothing rigid will fit, with capture methodology adaptable on the job, our smallest camera is 40mm in diameter |
No single column wins every row, and we’re not pretending otherwise. Laser scanning is excellent for collecting large amounts of data when static capture is suited to the project. Structured light has real strengths in controlled metrology environments. Stereo cameras are readily available and can cover the basics well..
But when you look at the rows that matter most for integrity campaigns — auditability, confined access, hardware flexibility, and whether the deliverable actually answers the engineering question — the picture shifts. And that’s before you consider what you might do when something goes wrong offshore and you need to capture data with whatever kit is on the vessel that day – “only single camera-photogrammetry can save us now..!”
Which method fits which job?
Technology comparisons in isolation only get you so far, because nobody scopes an integrity campaign by picking a sensor first. You start with a problem — something needs measuring, verifying, or monitoring — and work backwards to the method and deliverable that resolves it. The table below maps the scenarios we encounter most often against how each technology actually performs in that context, not in a sales presentation.
| Scenario | Laser Scanning | Structured Light | Stero Camera Systems | Stiil-based Photogrammetry |
|---|---|---|---|---|
| Retrofit / fit-up verification | Can provide geometry, accuracy at range degrades subsea, but commonly used for spool metrology | High accuracy in controlled setups, but limited working envelope makes full interface coverage difficult | Adequate for simple projects; struggles with low contrast subjects and capture of miniscule details. | Strong fit — sub-mm accuracy with in-scene verification, and the ability to work efficiently in both small and large worksites. "capable of detailing the dimensions of the template and hair-line cracks in the same dataset. |
| Crack growth / defect monitoring | Resolution typically insufficient for hairline cracks; comparing datasets across campaigns introduces alignment noise | Good surface detail at close range, but repositioning between campaigns is difficult to control precisely | Image resolution and non-strobed lighting of some units limits detection of fine defects; small fixed baseline constrains application of scale | Strong fit — tens of megapixels per image resolve micronic features where necessary and scale references provide a stable dimensional anchor during propagation analysis tasks |
| Confined / internal structures | Most units physically cannot access hawse pipes, caissons, or tight internal spaces | Small working envelope can suit some confined scenarios, but projector-camera geometry constrains positioning | Fixed rig either fits or it doesn't — no adaptation possible if the space is tighter than expected | Strong fit — a single camera and light can go where rigid systems cannot, with capture methodology adaptable on the job and miniaturised cameras |
| Emergency response (no specialist kit) | Requires the laser unit to be on board — rarely available unplanned | Requires dedicated hardware; not a realistic emergency option | Requires the manufacturer's rig on the vessel | Strong fit — certain worksscopes achievable with whatever camera is available, including kit already on the ROV or vessel |
| Legacy footage recovery | Not applicable — you can't retrospectively laser-scan from old video | Not applicable | Not applicable — requires original stereo pair from the manufacturer's rig | Viable in many cases, depending on footage quality, overlap, and whether enough spatial information exists to reconstruct 3D geometry |
| Large-area topside survey | Some systems offer near-real-time point clouds on deck | Near-real-time at close range | Can provide quick on-screen stereo measurements | Can be quickly augmented with a live-view system on request. |
Where does photogrammetry fall short?
We’re a photogrammetry company, so this section might seem counterproductive, but we’ve found over the years that being upfront about limitations earns more trust than pretending they don’t exist — and it saves everyone time.
Featureless, uniform surfaces.
Photogrammetry works by matching visual features across overlapping images. If the surface has no texture, no colour variation, and no distinguishing geometry — think a freshly coated flat steel plate with nothing on it — the software has very little to work with. There are workarounds (projected patterns, applied targets), but in truly featureless conditions, structured light or laser will give you a more reliable result. Luckily, this is almost never the case when working underwater!
Zero visibility.
If you genuinely cannot see the subject — silt-out conditions, heavy particulate, near-zero viz — then no optical method works, photogrammetry included. At that point you’re in acoustic territory (multibeam, sonar), which is a different discipline entirely and outside our scope. We’d rather say that plainly than let someone discover it offshore. Few turbid sites are permanently turbid however, and patience, tidal timing and local knowledge can help overcome this issue.
Massive standoff distances.
Unless you’re working in shallow water on a sunny Barbados day, you’ll need to take light with you, and light doesn’t do long haul underwater. If the object is more than three or four metres from your camera, you will struggle to get enough light onto the scene over that distance and can be forced into camera settings that degrade the quality of the output.
Reflective surfaces.
Reflective surfaces are a challenge to optical systems, but again are not a common occurence underwater, however we have seen instances where only a wire-brush was capable of removing certain types of marine growth and unnavoidably left a shiny surface in its wake. This brings us back to planning – single camera photogrammetry can overcome this issue by polarising the light, but only if the possibility was understood prior to mobilisation and the polarising equipment was mobilised.
None of these limitations are secrets, and most experienced subsea engineers have a reasonable feel for them already, but we’ve sat in enough tender debriefs to know that vendors who prioritise short-term gain over long-term trust tend to gloss over their method’s weaknesses and create problems downstream, sometimes costing the customer millions. We’d rather scope the work honestly and let the technology selection follow from the engineering requirement, even if that sometimes means recommending a method that isn’t ours.
How the use cases play out in practice
The table above gives you the overview, but the detail behind a few of these scenarios is worth expanding on because it illustrates why the technology choice alone doesn’t tell you much about the outcome.
Retrofit and fit-up verification
This is probably the highest-stakes dimensional work offshore. You’re fabricating a replacement or repair component to mate with geometry that only exists subsea, and if the measurements are wrong, you’re looking at rework, vessel standby costs, and potentially even a second mobilisation. What matters here is sub-millimetre accuracy that can be proven to the fabricator and to the client’s engineering team, plus a deliverable that goes straight into a fabrication drawing rather than a point cloud that someone else has to interpret. We’ve delivered over thirty first-time fit retrofit jobs on this basis, and the reason the hit rate holds up is that the accuracy is embedded in the dataset through in-scene scale references.
Confined and internal structures
That’s where we’ve seen the sharpest disconnect between what gets specified and what actually works on the day. Mooring chain hawse pipes, internal caisson structures, congested templates — if the worksite won’t accommodate a rigid stereo rig or a laser head on a pan-and-tilt, the options narrow very quickly. We’ve designed bespoke capture setups for diver and ROV led work in spaces where previous contractors had failed to get anything usable, and the common thread in those jobs was a lack of ability to ‘flex’ to the onsite situations.
Emergency response
A scenario most vendors can’t accommodate at all, because their service depends on their hardware being there. If the vessel is already offshore and something has gone wrong — a dropped object, a failed installation, an anomaly spotted during an unrelated operation — waiting for a specialist spread isn’t an option when standby costs run to six figures a day. Because single-camera based photogrammetry works with any camera, we can provide remote technical support and tailored capture procedures to the crew already on site. It’s not the textbook scenario and doesn’t cover the high-end detail, but it’s the one that saves the most money when it comes up.
Legacy footage recovery
This sits outside the traditional comparison entirely, because you’re not choosing a capture method — you’re working with whatever imagery already exists. Old ROV video, archived stills, footage from campaigns that were never planned for dimensional work. About 60% of the time, we can extract usable engineering data from this material, and the other 40% we’re upfront about. There’s no point pretending bad footage will give you good measurements, but when the footage is sufficient, the cost avoidance is significant — dimensional answers without putting a vessel in the water. What we’re most pleased about in this scenario, is that more and more engineerings now know they can transmit old inspection video and at least ask the question – ‘what can you make of this?’
What is dimensional assurance, and why does it matter?
Dimensional assurance means the 3D data you receive is accurate to a stated tolerance, that accuracy is independently provable after the fact, and the deliverable is something an engineer can act on (a fabrication drawing, a deviation report, an FEA-ready model) without needing a second opinion or a separate analyst. To remind, dimensional accuracy means different things in different scenarios. A 10mm tolerance to fit a flexible bung into a seatube is of course just fine, but it’s not going to cover you in glory if you use it to check propagation of a problem that expected to increase by 0.1mm per year! What about jobs that do have such a requirement – what if marine bio-life is eating your platform leg very slowly and your first scan happened to be at the low end of a 1mm tolerance and the subsequent scan at the high end? Are you really going to submit that report, knowing that it’s effectively telling the customer that the leg is healing itself? Thought not.
If you’ve read this far, you’ll have noticed that the technology comparison and the use cases both point in the direction of reliablity and assurance: what matters isn’t which sensor you deploy, it’s whether the output meets the required standard. We use the phrase deliberately, because “3D scanning” has become so broad it communicates almost nothing about the quality or usability of what you’ll receive. When buyers search for scanning services, they get a list of companies selling hardware time and the comparison collapses to price — which tells you nothing about whether the data will actually close out the work scope.
The analogy we keep coming back to (and we appreciate it’s not perfect, but it holds up well enough): we all know that a wedding photographer that offers to send video screen-grabs from an iphone is going to leave us very and irrepairably disappointed.
What should you ask your 3D scanning provider?
Whether you end up working with us or not, these are the questions that tend to separate providers who deliver dimensional outcomes from those who deliver datasets and move on:
- What engineering deliverable will I receive, and in what format?
- How is accuracy verified after the job — can I audit it independently, or am I relying on a pre-job calibration certificate?
- If the first capture doesn’t meet tolerance, what’s the recovery plan?
- Who interprets the data and produces the engineering output — is it the same team who captured it, or am I sourcing that separately?
- Is the data reliable and accurate enough for future propagation analysis to be similarly reliable and accurate?
If you get clear answers to those five questions, you’re probably talking to the right people regardless of what they call themselves. If the answers are vague, the service will be too.
We build our work around engineering problems, not scanning hardware. If you’ve got a dimensional question on an upcoming integrity campaign — or a dataset from a previous one that didn’t quite get you to the answer — give us a ring. Happy to talk it through.
FAQs
Accuracy depends on the method and conditions. Stills-based photogrammetry has demonstrated 0.05mm maximum error in Lloyd’s Register trials and 62-micron accuracy on North Sea campaigns. Laser scanning achieves sub-millimetre accuracy at close range but degrades with distance and turbidity subsea. Structured light offers tenths of a millimetre in ideal controlled conditions but has a small working envelope.
Stills-based photogrammetry is the strongest option for confined and internal structures. A single camera and lighting rig can access spaces where rigid systems physically cannot fit, with capture methodology adapted on the job. Most laser scanning units are too large, stereo camera rigs have fixed geometry that cannot be adapted, and structured light systems are constrained by projector-camera positioning.
Yes — stills-based photogrammetry works with whatever suitable camera is already available on the ROV or vessel. This makes it the only viable option for emergency response scenarios where waiting for specialist hardware is not practical. Laser scanning, structured light, and stereo camera systems all require their own dedicated hardware to be on board.
Dimensional assurance means the 3D data is accurate to a stated tolerance, that accuracy is independently provable after the fact, and the deliverable is something an engineer can act on directly — such as a fabrication drawing, deviation report, or FEA-ready model. Unlike generic 3D scanning, which may only deliver a point cloud requiring separate interpretation, dimensional assurance covers the full chain from capture to engineering output.
Photogrammetry is not ideal when real-time dimensional feedback is needed while the diver or ROV is still in the water, on featureless uniform surfaces with no texture or colour variation, in zero-visibility silt-out conditions where no optical method works, or at large standoff distances of 20 metres or more where laser scanning maintains accuracy more reliably.
