XGRIDS Pro Guide™ / Module 9: LCC Studio

9.3 Map Fusion in LCC Studio

Building a multi-segment model step by step: planning the segment limits, connecting segments correctly in the field, running the fusion at the workstation, inspecting the seams, and diagnosing the failures. Map Fusion is decided in the field, not at the workstation.

Section 1

Plan the Limits Before You Capture

Map Fusion stitches several scan segments into one model, for a site too large for a single battery or a single session. Every limit below is a field-planning decision. Exceed one and the fusion fails at the workstation, after the crew has left the site, with no way to recover except a return trip.

Limit	Specification
Segments per job	Up to 10.
Combined capture duration	200 minutes total across all segments in one job.
Per-segment capture duration	No limit. LCC Studio Map Fusion imposes no per-segment time cap, provided the job stays inside the segment count and the combined total.
Device capture ceiling	Each segment is still bounded by the scanner: 90 minutes on the L2 Pro and K2, 60 minutes on the PortalCam. XGRIDS additionally advises against PortalCam segments over 30 minutes, which produce drift and layering artifacts.
Device compatibility	All segments from the same device type: all L2 Pro, all K2, or all PortalCam. Types cannot be mixed in one fusion.
L2 Pro channel compatibility	16-line and 32-line captures cannot fuse together. The 32-line 120 m and 32-line 300 m modes can fuse.

No per-segment cap does not mean long segments are a good idea. Ten segments and 200 minutes is a lot of arithmetic freedom, and the guidance that actually governs segment length comes from elsewhere: your workstation's memory, the risk carried by any single long capture, and on an L2 Pro or K2 job, whether the same data will later go through the point cloud pipeline, which does cap each capture. Segment planning for the field is on 6.2 Scan Splitting.

Size the machine's RAM to the job before you capture. Fusion needs 64 GB minimum. A large job, 150 minutes or more of combined capture, run at high quality may need 128 GB. If the machine is short on RAM, run the job at Standard quality instead, which lowers the memory demand. The full hardware tier table is on 9.1 Pipeline Overview.

Section 2

Capture So the Segments Will Connect

Fusion succeeds or fails on how the segments were captured. Two segments connect only if they share a real link: an overlapping scanned area plus either RTK positioning or a matched control point.

A segment with neither RTK data nor a control point is rejected at import. LCC Studio will not accept it into the fusion job at all. Overlap and visual features alone do not admit a segment, and there is no way to add the link after capture. An uncontrolled segment must be re-captured, so verify RTK status or place control points during every segment, not just the difficult ones.

Overlap every segment with its neighbor

Every segment must share scanned ground with at least one other segment, directly or through the chain. An isolated segment cannot be fused. Rescan at least 50 ft of the previous segment's area, and aim for 65 to 100 ft of overlap. Choose an overlap zone with strong geometry: corners, columns, equipment, changes in depth. Avoid open plazas, long straight corridors, and smooth tunnels for the overlap, because a feature-poor zone gives the aligner nothing to lock onto.

Connect by RTK, or by control point

Method	Field procedure
RTK positioning	Keep RTK status Fixed for the whole of every segment. End the first segment, then start the next without moving the device. Rescan 50 ft or more of the previous area, then continue into new ground. LCC Studio applies the positioning automatically through Coordinate System Conversion, so there is no RTK toggle to set.
Control point	Use when RTK is unavailable or blocked (indoor, dense structure). A shared control point of the same name ties two segments together.

The control-point connection procedure

Run this at each seam between two segments captured without reliable RTK. One shared control point can tie two segments together. Three points spread across the overlap area is ideal, because more matched points give the aligner a stronger, drift-free join.

Scan the first segment. In the overlap area, choose up to three spots for shared control points.
At each spot, slowly lower the device to the ground with no sudden impact, add a control point in the field app, and name it in a simple sequence: fp01, then fp02, then fp03. Wait 15 seconds at each so the scanner captures a complete cloud around the point.
End the first segment.
Start the next segment. After initialization, revisit those same physical spots and add control points with the exact same names, fp01, fp02, fp03.
Rescan at least 50 ft back into the previous segment's area, then continue into new ground.
At the next seam, continue the sequence with fp04, fp05, fp06, and so on. Every shared point needs a matching partner of the identical name in the neighboring segment.

Control point names are case-sensitive and must match exactly across segments. A point named fp01 in one segment and FP01 in another reads as two different points. The connection fails silently, with no error, and the segment drops out of the fusion result. Decide a naming scheme before the crew starts and hold to it.

Pro Tip

Keep control point names lowercase with no dashes, underscores, or spaces, like fp01 and fp02. Simple lowercase names are faster to type on the device while you are scanning, and they sidestep the case-mismatch failure entirely.

PortalCam: place fusion points along an L-path

PortalCam marks the connection with on-device fusion points rather than survey control points. After scanning a segment, place three fusion points along an L-shaped path in the overlap area. For the next segment, walk that L-path in reverse and place three fusion points at the corresponding locations. Fusion points that mark the same physical spot must carry identical names. The full PortalCam procedure, including point spacing, the app steps, and the field sequence, is on 7.4 Map Fusion Collection Strategy.

Section 3

Connection Patterns Visual

This is a plan view, looking straight down at the site. Each colored box is one segment's coverage, and the hatched band is the overlap zone where the link lives. Step through the three valid patterns and the three failures before you plan a job.

Section 4

Correcting Control Points After Capture

A field naming error does not always mean a return trip. The Map Fusion interface has no per-marker editor, so correct control points in their files directly. In each affected scan folder, copy control_points.csv and control_points_latest.csv first, then make the change in both files so they stay consistent.

Fix a name. Edit the point name to match its counterpart in the other segment exactly, including case.
Remove a bad marker. A marker with an impossible distance to its neighbors, or two markers placed almost on top of each other, can fail the aligner even when the pre-fusion check counts it as present. Delete that marker's row from both files and rerun.

A green control point check confirms count, not quality. The pre-fusion dialog verifies that control points are present and counts them. It does not verify that their geometry agrees across segments. A fusion can still fail with every check green if the marker data is inconsistent. Treat the dialog as a completeness check, not a guarantee of a clean fusion.

Section 5

Run the Fusion in LCC Studio

The workstation half is a sequence: prepare the machine once, load the segments, preview before committing, set the parameters, clear the validation, then start and leave the machine alone. Each step below tells you what the setting or check does and what to choose.

One-time machine preparation

Set the storage paths before the first job. In Settings, under General, set the Project Path and the Capture Data Backup Path. Put both on an SSD and keep them separate from the software installation directory. SSD storage significantly improves processing efficiency.
Reserve disk headroom. Keep at least 2x the size of the capture project data free in the LCC data save directory. Running out of space mid-job interrupts or fails the reconstruction.
Confirm the hardware tier before committing the machine. RAM sizing for the job is in Section 1. The full Fusion-tier table (CPU, GPU, storage) is on 9.1 Pipeline Overview.

Create the job and load the segments

Click Create and select Map Fusion.
Batch upload the segments. LCC Studio reads each segment's RTK data and control point names and displays the device type it detects. Use Add Additional Capture Data to load segments in sequence, up to 10.
Upload from local storage or directly from the device over USB. USB mode reads local storage devices only (internal drives, SSDs, USB drives). Network storage such as OneDrive or a NAS is not supported as an upload source; copy the data to a local drive first.

Preview before you commit the machine

After the segments upload and before you start, read the Estimated Memory indicator. It is the software's own prediction of whether this machine can hold this job, and a red reading means it cannot. Fix that first, by dropping Reconstruction Quality to Standard, adding memory, or splitting the job, because a red estimate that gets committed anyway ends as an out-of-memory failure hours later with nothing to recover.

Then click Point Cloud Preview. LCC Studio checks its data indicators in turn; resolve anything it flags. Then click View Point Cloud to open the preview and inspect each segment's capture trajectory and rough point cloud. A few minutes here catches a coverage gap or a bad segment before the machine spends the night on a job that was never going to fuse. The preview navigates the same way as the LCC Scene Editor.

Set the parameters

Parameter	What it does	What to set
Reconstruction Quality	Fast, Standard, and Slow trade signal-to-noise ratio against time. Slow produces the highest quality and significantly increases VRAM consumption.	Standard for a first run. Move to Slow only after a Standard run succeeds and the deliverable justifies the added time and VRAM.
Maximum Gaussian Points	In fusion, the value applies per block, not to the whole model. LCC auto-adjusts block size to the scene, so the total point count is not capped by this setting.	Leave at or below 25M. Values above 25M have little effect on a fusion result.
Portability	On reduces model size and improves smoothness, especially on mobile. Off gives more realistic lighting at the cost of possible stutter.	On for models headed to web publishing or client viewing; Off for high-end local presentation.
Coordinate System Conversion	Replaces the old manual RTK toggle. The source system is identified automatically from the scan data. With valid RTK, the output converts to your target system; with abnormal RTK, the conversion is skipped and the model still builds; with no RTK, the option locks to None.	None for indoor work. A target system for outdoor RTK jobs that must align to GIS or BIM data.

Debug options and the remaining settings are documented on 9.8 LCC Studio Tools. Do not change them for a standard fusion.

Clear the pre-fusion validation dialog

Before the run starts, LCC Studio reports its checks: matching device model, matching LiDAR model, segment count, total scan time, RTK status, and control point count. Resolve any failed check before continuing, and remember from Section 4 that a green control point check confirms presence and count only, not geometric quality.

Start the run and leave the machine alone

Click Start to load the data. When loading completes, the My Models list opens.
On the project's card, click Start Reconstruction, then Confirm. The job enters the generation queue. If you are queuing several jobs, upload all their data before queuing begins.
Watch progress on the model card: the ring shows overall progress and the bar at the bottom shows the current stage. To pause deliberately, use the pause button on the card; resuming continues from the breakpoint.
Do not close LCC Studio, and do not run other GPU-consuming tasks on the machine while the job runs. Closing the application interrupts the job; on reopening, the card shows a failure status, and the card menu's Continue Generation or Restart Generation re-queues it. A large fusion runs overnight or longer, so schedule it and walk away.

Section 6

Inspect the Seams

A completed fusion is not a passed fusion. Open the result in the viewer and walk every boundary between segments before calling the job done or building deliverables on it.

Walk each seam at ground level. Look for doubled surfaces, offset walls, ghosted geometry, and density changes at the boundary. Clean seams are invisible; you should not be able to tell where one segment ends and the next begins.
Check the overlap zones specifically. The overlap is where alignment stress concentrates. A wall that appears twice, slightly offset, means the connection tolerance was exceeded in the field.
Verify a known dimension across a seam. Measure a feature that spans two segments. Significant deviation indicates the segments aligned locally but drifted globally.

A misaligned or doubled seam cannot be corrected in post. The connection data was insufficient at capture, and that segment must be rescanned with a proper link (Section 2). Plan the client schedule with this in mind: inspect seams the morning after the run, not the day the deliverable is due.

From a passed fusion, the model moves to the same downstream paths as any reconstruction: publish and author a tour on 9.12 Virtual Tour Workflow, or export through 9.7 Export Formats.

Section 7

When the Fusion Fails

Work the table top to bottom. The failure point tells you which field or workstation step broke.

Symptom	Cause	Fix
Segment rejected at import	The segment has neither RTK data nor a control point. LCC Studio does not admit uncontrolled segments to a fusion job.	Re-capture the segment with RTK Fixed or with shared control points (Section 2). There is no post-capture fix.
Segments will not load together	Mixed device types in one job, or L2 Pro 16-line data mixed with 32-line data.	Split into per-device jobs, or re-capture the mismatched segment on the matching device or channel count.
A segment is missing from the finished model, no error shown	Control point name mismatch, usually case (`fp01` vs `FP01`). The connection failed silently and the segment dropped out.	Fix the name in `control_points.csv` and `control_points_latest.csv` in that scan folder (Section 4) and rerun.
Fusion fails even though every pre-fusion check was green	The check confirms count, not quality. A marker with impossible geometry or two near-duplicate markers can fail the aligner.	Remove the bad marker's row from both control point files (Section 4) and rerun.
Estimated Memory shows red before the run	The combined capture is larger than this machine's memory can hold. The software is predicting the failure before it happens.	Drop Reconstruction Quality to Standard, add memory to the machine, or split the work into more than one fusion job. Do not start the run on a red estimate.
Job fails partway through the run	Insufficient RAM for the combined minutes, insufficient disk space, another GPU task competing, or LCC Studio was closed.	Check the total against the RAM guidance (Section 1) or drop Reconstruction Quality to Standard, confirm 2x disk headroom, close other GPU work, then re-queue with Continue Generation or Restart Generation.
Seam misaligned or doubled in the result	Connection tolerance exceeded in the field: weak overlap, feature-poor overlap zone, or degraded RTK during the seam.	Rescan that segment with 65 to 100 ft of feature-rich overlap and a solid link. Not correctable in post.

Pro Tip

Sketch the segment plan before the first scan: mark each segment, each overlap zone, and each shared control point name on a site plan or a phone photo. Fusion failures almost always trace back to a seam that was never captured or a name that did not match, and a two-minute sketch on site is far cheaper than a return trip.

Previous 9.2 Single Scene Next 9.4 Aerial-Ground Fusion