Remote Site Inspection With 3D Capture: Signing Off Without a Second Visit
A crack on the far side of a pole gets flagged three weeks after the crew has moved on. Remote site inspection with 3D capture turns the record into a space a reviewer can walk from a desk, so the follow-up question gets answered without a second trip.
A distribution technician climbs a pole, photographs a cracked crossarm, and files the report. Three weeks later the utility's engineer, reviewing the claim from a desk two states away, asks for the bolt pattern on the far side of the pole, the one angle nobody thought to shoot. Remote site inspection with 3D capture exists to answer exactly that follow-up without sending someone back up the pole: the technician captures the whole structure once, and the desk review happens against a space instead of a folder of stills.
The trip nobody budgeted for
A second visit rarely shows up as a line item until someone can actually bill for it. When a council can, the number turns concrete: the City of Edinburgh Council's 2026 fee schedule lists an additional site visit for a property inspection at £149, purely for sending someone back to look at what the first visit should have settled. Most firms absorb that cost quietly instead of invoicing it, which makes it easier to ignore and harder to fix.
The underlying problem is bigger than any one trip. FMI and PlanGrid, in a 2018 industry study, tied $31.3 billion of US construction rework to poor project data and miscommunication. A second site visit is a small, personal version of the same failure: a record made once that didn't hold the answer somebody needed later.
The volume makes it worse for utilities specifically. The American Society of Civil Engineers graded US energy infrastructure a D+ in its 2025 report card, down from a C- in 2021, citing an aging grid and rising demand. A backlog that size doesn't shrink by sending the same crew back to the same pole twice.
Photographs are the usual culprit. A folder of thirty stills answers the questions the technician thought to ask on site, and nothing else. A dispute that surfaces three months afterward, over whether a crack was already there or how far corrosion had spread, gets settled by whoever can still get back to the site. Not by whoever is actually right.
What remote site inspection with 3D capture actually changes
What changes isn't what the technician does on site. It's what happens to the record afterward. Instead of a folder of photographs chosen by whoever was holding the camera, the capture becomes a Gaussian Splat scene: continuous, photoreal surfaces reconstructed from the same photos, video, or 360 footage, opened in a browser rather than scrolled through as a gallery. A point cloud stays a cloud of measured points, accurate but sparse and hard to read for anyone outside the survey team. Reconstructed as a splat, the same data reads as a space, so a desk reviewer can move the camera to whatever angle actually answers the question in front of them, not just the angles the technician anticipated.
That's the part that removes the second visit. The reviewer isn't limited to the six photos that made it into the report. They're looking at the structure itself, and if the open question is the bolt pattern on the far side of the pole, they turn the camera and look.
For an insurer, this changes who touches a claim before it closes. A claims examiner who would normally wait on a re-inspection, or approve a settlement off a written description and a handful of stills, instead opens the same scene the field adjuster stood inside. The sign-off still requires a human decision. It just no longer requires that decision to wait on a second trip, or to be made with less information than the person who visited actually had.
The workflow
Capturing the asset
The person on site still has to be there once. What changes is what they capture. A walk-around video or a set of overlapping photos on an ordinary phone covers most poles, panels, rooftops, and equipment enclosures. A 360 camera works better inside a confined space such as a pump room or a switchgear cabinet, where a single sweep covers walls, ceiling, and floor in one pass. For an asset where a firm already owns or contracts LiDAR, point clouds in E57, LAS, LAZ, or PLY combine with the same imagery for a higher-precision result.
Uploading and processing
Processing happens in the cloud, with no GPU or install needed on the inspector's laptop. A single piece of equipment or a short walk-around reconstructs toward the fast end of a 10 to 120 minute range; a full site or a long pole line sits toward the slow end. Nothing about the wait requires anyone to stay put. The technician can be at the next stop on the route before the scene is ready to review.
Reviewing and marking findings
The output opens as a .spz, .ply, .splat, or .ksplat scene on a phone, tablet, desktop, or VR headset, with nothing to install on the reviewer's end. Inside the editor, hotspots carrying text, a photo, or a measurement note get dropped directly on the finding: the crack, the corrosion, the missing bolt. A desk reviewer working through a claim or a compliance check moves through the scene the same way they'd walk the site, except they can also line it up against a capture of the same asset from a previous cycle. Multi Splat aligns the two and renders the difference as a timelapse, so "has this gotten worse since the last inspection" is a rendered comparison rather than two people arguing over two separate photo sets.
Signing off and keeping the record
The scene, the hotspots, and the decision made against them become the record: shared as a link, embedded in a claims system, or pushed through the reconstruction API for a firm running inspections on a schedule rather than one at a time. The original photos or LiDAR scan stay exactly where they were, unchanged and available if a finding needs to hold up later. The splat scene is what the reviewer looked at to sign off. It isn't a replacement for the source data, and nobody should treat it as one.
Objections
Does a 3D capture actually count as an inspection record? It sits alongside the source photos or scan, not instead of them. What gets recorded is the scene, the specific hotspots raised against it, and the decision made, all timestamped to one capture. That's a fuller record than a report referencing three photos out of thirty, not a thinner one.
How do we know nothing was missed? This is the honest limit of the whole approach: a capture only shows what got captured. Coverage is still a field discipline, not something the reconstruction fixes after the fact. The gain isn't a guarantee against missed coverage, it's that a reviewer working from a wide, deliberate capture catches far more than one working from six chosen stills.
What about a site with no signal or hazardous access? Capture doesn't need a live connection. Photos, video, and LiDAR scans record locally and upload once the technician is back in range, so a remote substation or a rural pole line doesn't block the workflow, only the upload step waits.
Who has to learn a new tool? On the reviewer's side, nobody. Opening a link in a browser isn't a skill. On the field side, the shift is coverage, not software: shoot the whole structure instead of the six angles that answer today's question, which is closer to doing the same job more thoroughly than learning a new one.
| Site visit today | Remote review with 3D capture | |
|---|---|---|
| Time on site | One visit, plus a second if a question comes up later | One visit, comprehensive from the start |
| What comes back | A folder of photos chosen on the spot | A walkable scene covering the whole structure |
| Who can review it | Whoever can travel back, or join a screen share | Anyone with the link, from a desk |
| Answering a follow-up | Reschedule, or guess from the photos on hand | Turn the camera in the existing scene |
| What stays the record | The photo set, or the LiDAR file, on its own | The same source file, unchanged, plus the scene built from it |
FAQ
Can a 3D capture replace a physical inspection for compliance purposes?
Usually not on its own. Most compliance frameworks still require a qualified inspector to make the finding. What moves is where that inspector reviews the evidence from. The capture and its hotspots document what was seen and when, alongside the sign-off decision, but the regulatory requirement for a qualified reviewer doesn't change.
What happens to the raw photos or scan after the scene is built?
They stay exactly where they were, untouched. The splat scene is generated from them, not instead of them, so the source imagery or LiDAR file remains the record for any measurement or dispute that needs to go back to the original data.
How long before a reviewer can actually look at the capture?
Processing runs 10 to 120 minutes in the cloud once the file uploads, scaling with the size of the asset rather than the format it came in. A single enclosure or a short pole run is usually ready inside the hour.
Does the desk review need special software?
No. The scene opens in an ordinary browser on a phone, tablet, desktop, or VR headset. There's a link to click, not a client to install, which is the whole point of moving the review off the site and onto whatever screen the reviewer already has open.
Can this feed into a claims or asset management system instead of living in its own viewer?
Yes. A firm running inspections on a schedule rather than one at a time can push captures through the reconstruction API instead of uploading manually, and the resulting scene attaches to a claim or work order like any other document, rather than sitting in a link nobody files anywhere.
The crossarm from the opening didn't need a second technician back up the pole. The bolt pattern the engineer asked about was already sitting in the scene, three weeks old and still walkable from a browser two states away. That's the actual shift: not fewer inspections, but fewer trips back for a question the first visit could have answered if the record had been a space instead of a stack of photos.