5.1 Map Fusion Fundamentals
What Map Fusion is, when to use it, the hard limits that govern it, and how the three connection methods work.
What Is Map Fusion
Map Fusion is a post-processing mode in LixelStudio that combines multiple independent K1 or L2 Pro scan segments into a single unified point cloud dataset. Each segment is processed individually through SLAM optimization first, then the fusion algorithm aligns and merges them using shared spatial reference, either RTK coordinates, matching control point names, or both.
The reason Map Fusion exists is practical: a single SLAM session has limits. Battery life ends. RAM fills. A 20-floor building cannot be scanned in one unbroken session. Map Fusion is the mechanism that allows you to break a large project into manageable segments during collection and reassemble them into one coherent dataset in post-processing.
Map Fusion is not a stitching tool for overlapping redundant data. It is an alignment tool for sequential scan segments that together cover a complete site. The segments need to share spatial reference at their connection zones, that reference is what allows the algorithm to lock them together accurately.
When to Use It
5 situations make Map Fusion the correct approach rather than attempting a single long scan.
Projects Exceeding Battery Life
The L2 Pro and K1 each provide up to 90 minutes of scanning time per battery charge. Large facilities, warehouses, campuses, and multi-floor commercial buildings require more time than a single battery supports. Rather than scanning until the battery dies, plan segment breaks at logical transition points and execute Map Fusion to join them in post-processing.
Projects Exceeding RAM Capacity
LixelStudio holds the entire scan trajectory in memory during SLAM optimization. On a 64 GB machine, scans longer than approximately 30 minutes risk failure. On 128 GB, the ceiling is approximately 60 minutes. If your project scope exceeds your hardware, Map Fusion is the solution: process shorter segments that fit in RAM, then fuse the results.
Multi-Floor Buildings
Moving between floors on stairways creates SLAM stress, narrow geometry, repetitive structure, brief feature interruption. Breaking at floor transitions and using Map Fusion produces more reliable results than attempting to carry a single session through multiple stairways.
Long Linear Environments
Tunnels, mine galleries, and corridors exceeding 1,600 ft (500 m) are not suitable for a single Narrow Mode session. The correct approach is to break these into segments of 1,600 ft or less and use Map Fusion with relative control points to join them. Field execution for this scenario is covered in detail in the Field Collection section.
Recurring Scan Projects
Facilities that require periodic rescanning on a scheduled cycle (construction progress monitoring, facility management, asset tracking) use Map Fusion to merge new scan segments with existing baseline data. Only the changed zones need to be recaptured. The new segments are fused with the unchanged baseline through shared control points.
This scenario requires control point infrastructure to be planned into the baseline scan from day 1. Control points cannot be added to a completed scan after the fact. If the baseline has no shared control points with the future rescan zones, fusion is not possible. See the Control Point Planning section in 5.2 Pre-Project Planning for permanent control point requirements.
Segments captured over several weeks or months can be integrated into the same Map Fusion project, provided that data collection requirements are met: sufficient overlapping areas, GCPs present, and the scene remains relatively static between capture sessions. Significant changes between sessions (moved furniture, construction activity, seasonal vegetation changes) cause visual matching errors that produce blurring or ghosting in the fused result.
Use the same firmware version for all segments in a fusion project. Mixing firmware versions between data collection sessions can introduce processing inconsistencies. If you need to reprocess, you must reprocess the entire dataset. Partial updates are not supported. For example, if you have a fused model from 10 segments and are unsatisfied with segment 10, you must re-collect segment 10 and then reprocess all 10 segments together to generate the final model. You cannot replace only the affected segment in an existing fused output.
Loop closure still matters within each segment. Map Fusion aligns segments to each other, but it cannot fix drift that already exists inside a segment. Each individual scan must be executed with proper loop technique before fusion adds any value.
Hard Limits
These are not recommendations. They are system constraints. Exceeding them causes fusion to fail or produce unusable output.
Overlap placement determines fusion reliability. Selecting overlap zones with rich, stable surface features, furnished rooms, structured intersections, equipment areas, gives the algorithm strong matching geometry. Placing overlap zones in hallways, stairwells, or near reflective surfaces gives it weak or ambiguous geometry. Poor overlap placement is the most common cause of fusion failure on otherwise good data.
Complete overlap also causes failure. If 1 segment's scan path is entirely contained within another segment's scan path, the algorithm cannot determine the correct spatial relationship between them. Each segment must cover some unique area, with the overlap zone serving as the connection bridge, not as the entire content of the segment.
Connection Methods
Map Fusion uses 1 of 3 methods to establish alignment between segments. The method you use determines whether the fused output has global coordinates, relative coordinates, or no georeferencing at all. This decision must be made before collection begins, it cannot be applied retroactively.
RTK-Based Fusion
Each segment carries valid RTK data acquired during scanning. The fusion algorithm uses the shared coordinate system embedded in each segment's RTK record to align them.
- Requires Fixed RTK status during scanning, not Float or Single Point
- Each segment must independently meet RTK validity thresholds
- Any segment with valid RTK pulls global coordinates into the fused output
- Best choice when outdoor RTK coverage is reliable across all collection windows
Control Point Fusion
Shared physical locations are marked with identical names across consecutive segments. The algorithm matches same-named points between segments to establish alignment.
- Each consecutive segment pair must share at least one identically named point
- Point names must match exactly, every character, every capital letter
- Device must remain on the ground when marking control points
- Rescan 50 ft (15 m) of overlap after marking each control point
- Output has relative coordinates unless RTK is also present
Hybrid Fusion (RTK + Control Points)
Some segments carry RTK; others are connected through shared control points. The algorithm uses whatever reference is available per segment pair. This is the most common approach for large mixed-environment projects: outdoor areas on RTK, interior areas on control points.
- Any segment with valid RTK pulls global coordinates into the result
- Control point connections fill the gaps where RTK was unavailable
- For indoor-outdoor transitions, place 3 relative GCPs in an L-shaped distribution at the transition zone. This distribution gives the algorithm non-collinear geometry to resolve both horizontal and rotational alignment
- Both RTK and relative GCPs are used simultaneously in this mode
The 4 Valid Connection Patterns
Official documentation defines which configurations the fusion algorithm can resolve.
- All segments have RTK
- No RTK: consecutive pairs share identically named control points
- Segment 1 has RTK; segments 1 and 2 share control points; segments 2 and 3 share control points
- Segments 1 and 2 have RTK; segments 2 and 3 share control points
Every segment must be linkable to at least one adjacent segment through one of these patterns. A segment with no connection to its neighbors cannot be fused.
What Comes Out
The output of a successful Map Fusion job is a single unified point cloud that spans all input segments. If any segment carried valid RTK, the output has global coordinates. If only control points were used, the output has relative coordinates, internally consistent geometry with no absolute position.
The fused result exports from LixelStudio in E57, LAS/LAZ, RCP, or PLY format. The same export selection and quality considerations that apply to single-scan outputs apply here. Regardless of how many segments contributed, the final output is one dataset.
Processing time for Map Fusion is substantially longer than single-scan processing, and hardware requirements are higher. The next sections cover pre-project planning, field collection execution, and the full processing workflow.
This module covers Map Fusion in LixelStudio (point cloud output). Map Fusion also exists in LCC Studio for 3D Gaussian Splat output from K1, L2 Pro, and PortalCam. The field collection requirements are the same, but the processing workflow differs. See 9.3 Map Fusion in LCC Studio for the 3DGS processing procedure.
©2026 Alpine Reality Capture LLC • XGRIDS Pro Guide™ • Site Disclaimer