Sometimes a capture is reconstructed in two or more segments rather than as one continuous model. Here’s why that happens, how the segments are joined, and what it means for your data.

You may end up with a capture flagged and are wondering what that means - and that is why you are here. The automated data flagging can be found at: Automated Data Flagging

Why a capture splits

The LOOQ device takes a set of images roughly twice per second as you walk. Reconstruction works by matching what it sees in one set against the next, building a chain from the start of your walk to the end. That chain is what holds the capture together in one piece.

Sometimes the chain breaks. Reconstruction reaches a point where it can't connect what it's looking at to anything it has already built, so it closes off that piece and starts fresh. The result is two or more segments — each a correct reconstruction of part of your walk, but with no built-in relationship to the others.

The conditions that break the chain are the same ones that make a capture hard to reconstruct at all:

• Nothing distinctive to match on. Blank walls, windows, plain grass, or scenes with no vertical features.
• Dense vegetation or tall grass. Repeating patterns look the same from one image to the next and confuse the matching.
• Too close to what you're capturing. Inside about 1 m (3 ft) — narrow walkways, walls, hedges — consecutive images don't overlap enough. The same happens when the operator is in frame, blocking the side cameras.
• Moving too fast. Above a brisk walking pace the qCam can't capture often enough to keep images overlapping, and they blur.
• Low light. Dawn, dusk and night force the camera to raise exposure and gain, which adds noise and blur.
• Rain, snow, or wind. Drops or fogging on a lens, a white-out where sky and snow blend together, or vegetation moving between images.
• Water. Reflections and a moving surface leave nothing static to match against.

A capture that split used to fail outright — even when most of the walk had reconstructed perfectly. That good data was thrown away. It isn't any more.

Full guidance. These are the conditions most likely to split a capture. For the full set, see Unsupported Capture Environments.

How the segments are joined

Each segment comes out of reconstruction in its own arbitrary position, orientation, and size, with no visual link to the others. But every image in every segment has a GNSS position attached, recorded by the device when it was taken. That's the common reference.

Each segment is fitted onto its own GNSS positions — solving for where it sits, which way it faces, and how large it is in real metres. Once all the segments have been fitted, they share one projected coordinate frame and can be combined into a single point cloud.

Left: three segments, each in its own arbitrary position, rotation, and size, with no shared reference. Right: after each is fitted onto its own GNSS positions, all three share one projected coordinate frame and form a single continuous route.

Segments come out of reconstruction with no shared reference. Fitting each one onto the GNSS positions recorded for its own images puts them all in the same projected frame, so they can be combined into a single capture. A seam is where two segments meet.

From there, the capture is processed exactly like any other: dense reconstruction, panoramas, ortho, and exports all run normally, and you receive one deliverable.

This depends on good GNSS. The device's positions have to be accurate to a few centimetres for the fit to work — in practice, a properly logged base station and a clean PPK solution. Without one, there's no shared reference to join the segments on, no matter how good the imagery is.

What it means for your data

Inside a segment, nothing changes. Each segment is a full-strength reconstruction. Measurements between two points in the same segment are as accurate as they've always been.

Across a seam, accuracy is GNSS-limited. A seam is where two segments meet. Features on opposite sides of a seam are related to each other through GNSS rather than through the imagery, so that relationship is centimetre-level rather than the tighter accuracy you get within a segment. For a critical measurement that crosses a seam, check it against a known control point.

Seams can show slight doubling. Where segments overlap physically — you walked back past a pole you'd already captured — surfaces in that overlap can appear slightly doubled. Minor doubling at a seam is expected, not an error.

When the segments can't be joined

If the GNSS isn't good enough — no base station, a failed PPK solution, heavy canopy, or tall buildings throughout the walk — the capture still isn't failed. You receive the largest single segment instead.

That's a correct, usable point cloud of the longest continuous stretch of your walk. It just won't cover the rest. If a capture comes back covering noticeably less ground than you walked, this is the most likely reason, and GNSS is the first thing to check.

Checking a segmented capture

1. Coverage. Does the cloud cover the ground you actually walked? A large shortfall points to the case above.
2. Position. Confirm the cloud lands where it should against a control point or basemap. This is the most valuable single check.
3. Seams. Look where coverage transitions for doubled surfaces, a step in the ground, or a break in a curb, fence or conductor.
4. Features that cross a seam. A pole or span running from one segment into the next should read as one continuous object.

Contact support if you see. The cloud in the wrong place, a step of more than a few centimetres at a seam, a section rotated relative to the rest, or coverage well short of your walk.

Avoiding a split in the first place

Segmented reconstruction is a recovery. A capture that stays whole is still the better outcome. The full rules are in Unsupported Capture Environments — these are the ones that most often keep a capture in one piece:

• Keep at least 1 m (3 ft) between the qCam and anything you're capturing — walls, vegetation, people — and stay out of your own frame so the side cameras aren't blocked.
• Capture on foot, up to a brisk walking pace. On a vehicle mount, stay under 10 km/h (6 mph) and keep the vehicle out of frame. Don't capture from a boat.
• Capture in good light. Not dawn, dusk, or night. If the images look green, stop and come back when there's more light; use flash mode in small enclosed spaces.
• Give the scene something to hold onto — brick, concrete, roadway markings, poles. Move slowly and continuously rather than in stops and starts.
• Clear tall grass and vegetation beforehand rather than pushing it aside with hands or feet as you go.
• Wait out bad weather. In light rain, wipe the affected lenses regularly.
• Stay outdoors. The system relies on GNSS from the qCam's antenna, which is weak under heavy canopy and beside tall buildings, and absent indoors. If you have to pass through a GNSS-denied stretch, begin and end the capture where GNSS is strong.

Frequently Asked Questions

Does a segmented capture cost me accuracy?

Not within a segment. Across a seam, the relationship between features comes from GNSS rather than the tighter accuracy inside a segment.

How many segments is too many?

There's no count that makes a capture unusable. What matters is GNSS quality and how much of the walk was reconstructed.

Why is my capture smaller than what I walked?

Most likely the segments couldn't be joined, so you received the largest one. Check the GNSS for that capture.

Can it be re-processed?

Yes. If the GNSS solution improves — a corrected base station log, a re-run PPK solution — the capture can be re-processed and may join successfully. Contact LOOQ support with the scan ID.