5.4 Map Fusion Processing

Hardware Requirements

Map Fusion is more computationally demanding than single-scan processing. Verifying hardware before starting a long fusion job prevents a failure at hour 30 of a 65-hour run.

Recommended specifications for Map Fusion: 16-core processor (AMD Ryzen 9 9950X or equivalent), 64 GB DDR5 RAM with 96 to 128 GB recommended for larger datasets, NVIDIA RTX 3090 with RTX 4090 recommended for optimal performance, and an NVMe SSD with workspace equal to approximately 5 times the raw data size.

Project Setup

Setting up a fusion project correctly at the start prevents configuration errors that require restarting the entire job.

1. Transfer All Segment Data to the Processing MachineCopy all segment folders from the scanner to a dedicated project folder on an NVMe SSD. Keep each segment in its own folder, named according to your planned naming convention. Verify all files transferred completely before beginning processing.
2. Open LixelStudio and Create a Map Fusion ProjectSelect Map Fusion as the processing type. Do not create a single-scene Project Processing job and attempt to add multiple segments; Project Processing and Map Fusion are separate pipelines, and the type is set at creation.
3. Add Maps in Collection OrderClick Add Map and import each segment in the order it was scanned. The list displays each map's Point Name, File Path, and Mode. The sequence matters for control point matching; the algorithm processes connections between adjacent segments based on load order. Remove a single map with the trash icon, or Clear to remove all.
4. Set the Per-Map ModeUse the Mode dropdown in the import list to match each segment's capture scene: Standard for general scenes, Robust for segments needing higher processing stability, Narrow Space for tunnel segments. A mixed project can run different modes per segment.
5. Upload Panoramic Video Files for Panoramic Camera SegmentsIf segments were scanned with the panoramic camera, upload the matching panoramic video file for each segment at this stage. Segments captured without the panoramic camera skip this step automatically.
6. Set the Base Map and Output PathChoose Auto or designate a Base Map (see Configuration below for when to designate). Set the output path, and optionally check Auto-load project after processing and Point Cloud Segmentation. Confirm free disk space covers the workspace requirement before starting; running out of RAM or disk fails the job mid-processing, not at the start. If memory is tight, reduce the number of segments per fusion job and merge the outputs downstream.

Configuration

The defaults are appropriate for most projects. The settings below are the ones to understand when the defaults are not the right choice. Basic Parameters and Coloring work the same way in Map Fusion as in single-scan Project Processing.

Processing Phases

Map Fusion processing runs in three sequential phases. Understanding what each phase does helps you interpret progress indicators and diagnose where a failure occurred.

1. Individual SLAM Optimization (Per Segment)Each imported segment runs through the standard SLAM optimization pipeline as if it were a single-scan project. Drift correction and camera-LiDAR fusion happen here. A segment with collection problems may fail this phase; the fusion algorithm cannot proceed until all segments complete their individual optimization.
2. Inter-Segment AlignmentThe algorithm computes the spatial relationship between every adjacent segment pair using RTK coordinates, matching control point names, or both. This is where the connection method you selected during planning takes effect. Misaligned segments at this phase indicate poor overlap geometry, mismatched control point names, or insufficient shared coverage.
3. Final Unified Output GenerationThe fully aligned segments are merged into a single trajectory and point cloud dataset. Coordinate system transformation (if RTK is present) and any post-alignment refinements are applied. The output is the single fused point cloud ready for quality review and export.

Quality Review

After processing completes, review the fused result before exporting. A quality review that takes 15 minutes catches problems that would waste hours of downstream work in Revit or AutoCAD.

Segment Boundary Inspection

Navigate to each segment boundary, the area where two segments join, and inspect the geometry. A well-aligned fusion shows seamless continuation of geometry across the boundary. Misalignment shows as a step, doubling of surfaces, or geometry that does not line up between the two sides of the boundary. If misalignment is visible, the connection at that boundary needs to be investigated before the result is used.

Accuracy Report and RMSE

Review the accuracy report generated after processing. The RMSE values at each control point connection show how closely the algorithm achieved the theoretical alignment. High RMSE at a specific connection indicates a problem with that segment pair's connection data: poor overlap, a control point with inconsistent position between takes, or an RTK validity issue.

Coverage Completeness

Verify that the fused output includes data from all segments and that there are no unexpected gaps at the transition zones. A segment that failed during individual SLAM optimization may be absent from the fused result; the final dataset will be missing the territory that segment was supposed to cover.

Export

The fused result exports from LixelStudio as a single dataset in LAS, LAZ, PLY, or E57. For structured E57, use the built-in LAS-to-E57 tool. For Autodesk workflows, the built-in LAS-to-RCP converter handles up to 10 LAS files per conversion and produces RCP for AutoCAD and ReCap 2020 and later, no ReCap Pro license required. The format selection follows the same decision logic as single-scan exports.