6.4 Aerial-Ground Fusion
Aerial-Ground Fusion combines drone imagery with ground-level L2 Pro scan data to produce a complete model covering both perspectives. The L2 Pro captures ground-level geometry and detail that the drone cannot see. The drone covers building exteriors, rooftops, and large outdoor areas that ground scanning cannot efficiently reach.
What Aerial-Ground Fusion Is and When It Applies
Aerial-Ground Fusion is L2 Pro only. The K2 and PortalCam do not support this workflow. If your project requires a fused aerial-ground deliverable, plan to use an L2 Pro for the ground component and a separate mapping drone (such as a DJI Matrice with a Zenmuse P1 payload) for the aerial component.
Aerial-Ground Fusion is not the same workflow as L2 Pro Drone Mode covered in section 6.3. In Drone Mode, the L2 Pro is physically mounted on the drone and collects LiDAR data from the air. In Aerial-Ground Fusion, the L2 Pro stays in your hands as a ground scanner and a separate mapping drone collects oblique photographic imagery from the air. The 2 datasets are aligned in software after capture.
The output format also differs depending on which software processes the data. LCC Studio produces a 3D Gaussian Splat from the combined dataset. LixelStudio produces a merged point cloud. The choice between them depends on your deliverable, not on the field collection method. Both pipelines require the same field data collected the same way.
LCC Studio Pipeline
Output: 3D Gaussian Splat (3DGS). Photorealistic rendered model viewable in LCC Viewer, suitable for walkthroughs, client delivery, and web publishing.
Ground device: L2 Pro.
Ground data input: Raw scanning data only. The output of a completed Map Fusion project cannot be used as ground data. Maximum 10 ground segments, total duration under 200 minutes (same constraints as standard Map Fusion).
Aerial data: 100 to 10,000 drone images. JPG/JPEG only. Minimum resolution 1024×768, consistent across all images.
Fusion method: Single aerial image set uploaded alongside the L2 Pro ground scan.
Processing time: 24 to 48 hours or more for large sites. Hardware minimum: 16-core CPU, 96 to 128 GB RAM, RTX 4090.
LixelStudio Pipeline
Output: Georeferenced merged point cloud in E57, LAS/LAZ, or RCP format. Suitable for BIM, survey deliverables, measurement, and coordination with other point cloud datasets.
Ground device: L2 Pro.
Aerial data: Drone imagery organized in a Drone/ subfolder alongside the Lixel/ project subfolder. XGRIDS Pose tool (extract_preview_poses) must be run on the drone data before LixelStudio import.
Fusion method: Both the Drone/ and Lixel/ subfolders are imported into LixelStudio as a single project. LixelStudio merges the point clouds using the RTK tracks from both datasets.
Processing time: Varies by data volume. Typically longer than a standard ground-only scan due to the aerial data processing overhead.
LCC Studio Aerial-Ground Fusion accepts raw scanning data only as ground input. You cannot process a Map Fusion project first and then feed the fused output into an Aerial-Ground Fusion project. The ground data slots accept raw, unprocessed scan segments directly from the device. Ground segments follow the same constraints as standard Map Fusion: maximum 10 segments, total duration under 200 minutes. Plan your ground collection to fit within these limits for a single Aerial-Ground Fusion job.
The field collection procedure is largely the same regardless of which software pipeline you use. The difference is what you do with the drone images after collection: organized takeoff/landing folders for LCC Studio, or the Drone/Lixel folder structure with the Pose tool for LixelStudio. Plan your folder organization before the flight. Reorganizing a large drone image set after the fact is tedious and error-prone.
L2 Pro Drone Mode LiDAR data cannot be used as the aerial component in aerial-ground fusion. No workflow exists in LixelStudio or LCC Studio that accepts both an aerial L2 Pro drone mode point cloud and a ground L2 Pro point cloud as inputs for a combined project. The aerial-ground fusion pipelines in both applications require photographic drone imagery as the aerial component, not aerial LiDAR data. If your project requires both aerial coverage types, plan for 2 separate drone configurations on the same platform. The first flight uses a mapping camera such as the Zenmuse P1 to collect the photographic imagery that the aerial-ground fusion pipeline requires. That image set is what gets combined with your ground scan data in LCC Studio or LixelStudio to produce the fused model. The second flight swaps the camera payload for the L2 Pro drone mounting bracket and collects an aerial LiDAR scan. That scan processes independently in LixelStudio as its own point cloud in E57, LAS/LAZ, or RCP format, suitable for elevation modeling, roof geometry measurement, terrain surfaces, and coordination with other point cloud datasets. The 2 deliverables serve different purposes and do not combine with each other in processing.
Project Types That Benefit from Aerial-Ground Fusion
- Campus facilities and multi-building complexes where exterior context and interior detail are both required
- Construction sites needing periodic full-site documentation including ground-level progress and aerial site overview
- Heritage documentation of buildings with significant exterior architectural detail that cannot be fully captured from the ground
- Large outdoor areas with connected interior spaces, such as industrial yards leading into warehouses
- Any project where a roof, upper facade, or terrain above the scanner's reach is part of the deliverable
Hard Requirements Before You Start
Aerial-Ground Fusion has non-negotiable prerequisites that apply to both the ground and aerial data collection. A project that does not meet these requirements will fail during processing. There is no workaround in the software for missing RTK data or incorrect coordinate system configuration.
Drone Recommendations
The best aerial platform for Aerial-Ground Fusion is a DJI Matrice 300 or Matrice 350 equipped with a DJI Zenmuse P1 camera payload. The Zenmuse P1 is a full-frame mapping camera, not a drone itself. It mounts to the Matrice platform and provides significantly higher image quality and geometric accuracy than lighter consumer-grade alternatives.
If your operation already runs a Matrice 300 or 350 for L2 Pro Drone Mode, you can swap the L2 Pro mounting bracket for the Zenmuse P1 payload and use the same drone for the aerial photography phase of a fusion project. No additional drone platform is required.
Site Planning and Fusion Point Selection
Before arriving on site, identify 4 to 5 fusion point locations using satellite imagery or aerial photos. These are the physical locations where the drone will take off and land, and where the ground scanner will perform its fusion-specific collection routine. These points are what LCC Studio uses to align the 2 datasets in three-dimensional space.
What Makes a Good Fusion Point
A fusion point location must be accessible to both the drone (for takeoff and landing) and the scanner (for ground-level scanning). Beyond that, it needs to meet all 4 of these criteria:
- Open area: The drone needs a clear vertical path for takeoff and landing. Overhead obstructions like trees, power lines, or building overhangs will prevent the required vertical image sequence from being captured correctly.
- Rich surface features: The fusion algorithm matches geometry visible in both the drone images and the ground scan. A blank pavement square with no features provides almost nothing to match against. A paved area with planters, signage, bollards, or building corners is much more useful.
- Stable and unchanging between visits: If the ground scan happens on Tuesday and the drone flight happens on Thursday, the fusion point area must look the same in both datasets. Parked vehicles, staging materials, or any moveable objects that might not be there both days will degrade the fusion result.
- Accessible to the scanner from multiple angles: The scanner needs to circle the fusion point area to provide 360-degree ground-level coverage that the drone can match against from overhead.
Distribute fusion points evenly across the site. 4 to 5 points should span the full extent of the project area. 2 points clustered at one end of a large site will not provide the geometric constraint needed for the far end of the dataset to align correctly. Sketch a site plan and place fusion points so that no part of the project boundary is more than roughly 50 to 100 meters from the nearest fusion point.
Ground Collection: L2 Pro Method
For Aerial-Ground Fusion, the L2 Pro ground scan follows a standard RTK-enabled scan procedure with one specific addition at each fusion point: the scanner must complete 2 loop closures within 2 meters of the takeoff and landing location. This gives the SLAM algorithm dense, well-constrained ground-level geometry precisely at the area the drone will be photographing from directly above.
Verify RTK is Fixed before starting the scan
In LixelGO, confirm satellite status is Fixed (not Float, not Single Point) and that you have at least 10 valid satellites. Set the coordinate system to WGS84. The drone must also be set to WGS84. Do not begin scanning until RTK is Fixed. L2 Pro RTK uses the optional add-on module.
Scan the site following standard L2 Pro technique
Scan the full project area normally. Maintain RTK Fixed status throughout as much of the scan as possible, particularly in and around the fusion point areas. If RTK signal is lost under cover or inside structures, the fusion point areas must be covered with Fixed RTK.
At each fusion point, complete 2 loop closures within 2 meters of the location
When your scanning path brings you to each fusion point location, make 2 distinct loops around the point, each staying within 2 meters of the takeoff/landing spot. Walk the loops at different heights if the area allows. These loops create the dense, geometrically constrained ground data at the exact location the drone will capture from above.
If you arrive at the fusion point from one direction and leave in the same direction without completing the loops, the scan data at that point will not be sufficiently constrained for reliable fusion alignment.
Continue through all fusion points and complete the full scan route
Visit all planned fusion point locations during the scan. Complete the full site coverage and close the overall scan loop before stopping. Stop the scan in LixelGO and wait for the solid green LED before powering off.
Drone Flight and Aerial Data Collection
The drone flight has 2 distinct phases that must both be completed at every fusion point location: the main grid mission over the site, and the takeoff/landing image sequences at each fusion point. The grid mission provides the broad aerial coverage. The takeoff/landing sequences provide the visual bridge between the aerial and ground perspectives. Without both, the fusion will fail or produce degraded results.
Main Grid Mission
Set the drone coordinate system to WGS84
On the DJI remote controller, confirm the coordinate system is WGS84 before setting up the flight mission. This must match the ground scanner's coordinate system setting.
Plan a smart grid flight pattern over the survey area
Use the DJI controller's smart grid function. Frame the survey area on the controller map to define the flight boundary. A tic-tac-toe pattern (2 perpendicular sets of parallel flight lines) provides the most complete coverage for Aerial-Ground Fusion.
Set capture mode, GSD, gimbal angle, altitude, and overlap
Use oblique capture mode with the gimbal angle set to 45 degrees. Set GSD (ground sampling distance) to 1 centimeter or below. Set flight height above the tallest building in the survey area to ensure clearance. In advanced settings, set both side overlap and forward overlap to 85 percent.
For large areas with complex structures, plan 1 to 2 additional altitude passes above the main altitude. The height difference between adjacent passes must not exceed twice the lower altitude.
Execute the grid mission and monitor continuously
Tap Start. The drone will fly the planned route automatically. Monitor the drone throughout the flight. If conditions change (wind, unexpected obstacles, airspace alerts), pause the mission and resume when it is safe to continue. Do not leave the drone unmonitored during the mission.
Takeoff and Landing Image Sequences
After the main grid mission, the drone must visit each fusion point location and capture a continuous image sequence from ground level to flight height. This is the most critical and most frequently skipped step in Aerial-Ground Fusion. A grid flight without these sequences will not fuse reliably with the ground scan.
Plan at least 3 to 4 takeoff/landing sequences distributed throughout the scanning path. For larger sites, plan at least one sequence every 50 to 100 meters of the ground scan route. Each sequence must provide continuous coverage from approximately 1.5 meters above the ground up to a partial altitude, with adjacent image overlap of 85 percent or greater. Insufficient sequences or image overlap below 85 percent causes LCC Studio to fail alignment between the ground scan and aerial imagery. The resulting model will have missing sections or gross misalignment that cannot be corrected without recollecting the data.
Fly to the first fusion point and descend to approximately 1.5 meters above ground
After the grid mission is complete, navigate the drone to the first fusion point. Descend to approximately 1.5 meters above the ground and position the drone so the camera is pointing at fixed, stable features that were also visible to the ground scanner at that location.
Capture 30 to 40 photos while ascending steadily from ground level to flight height
Begin capturing images at approximately 1-second intervals while slowly and steadily ascending from the 1.5-meter starting position up to the full flight altitude used during the grid mission. The sequence must be continuous with no large position jumps. The lowest images must overlap spatially with the ground scanner's trajectory. The highest images must overlap with the grid flight imagery.
Shoot from at least 3 different horizontal directions at each fusion point to ensure sufficient angular coverage for the fusion algorithm.
Keep adjacent image overlap at or above 85 percent throughout the sequence
The value of the takeoff/landing sequence is the visual continuity between ground level and flight altitude. If there are large gaps between images (the drone moved too fast, or too few images were captured), the transition between the ground and aerial perspectives will be broken and the fusion alignment will fail at that point.
Repeat at every fusion point before landing
Visit each planned fusion point location in sequence and capture the full takeoff/landing image sequence at each one. Keep the images from each fusion point in separate folders. Do not mix images from different fusion point locations into a single folder.
Folder Organization Before Processing
How you organize the drone images after the flight determines how you will import them in LCC Studio or LixelStudio. Organize the folders before you sit down to process. Doing it during an active processing session introduces errors.
L2 Pro Method (LixelStudio)
LixelStudio expects a parent folder with 2 subfolders: Drone for the aerial imagery, and Lixel for the L2 Pro scan project folder. Before importing into LixelStudio, the XGRIDS Pose tool (extract_preview_poses) must be run on the drone data to prepare the camera pose files that LixelStudio requires.
├─ Drone/ # all drone image data goes here
│ └─ (all drone images, organized by flight folder)
└─ Lixel/ # L2 Pro project folder goes here
└─ (project folder from the device)
Place the L2 Pro project folder in the Lixel subfolder
Copy the project folder from the device into the Lixel/ subfolder of your project directory. Do not alter the internal structure of the project folder.
Place all drone images in the Drone subfolder
Copy all images from the drone's SD card into the Drone/ subfolder, preserving the DJI-generated folder structure. This includes the grid flight folders and the takeoff/landing sequence folders.
Run the XGRIDS Pose tool on the drone data
The XGRIDS Pose tool (extract_preview_poses application, provided by XGRIDS) processes the drone images and generates the camera pose data that LixelStudio needs to align the aerial imagery with the LiDAR point cloud. Run this tool before opening LixelStudio. You can use the included visualization tool to verify that all camera poses have been correctly extracted before proceeding.
Import the parent folder into LixelStudio and configure processing
Open LixelStudio and navigate to the Aerial-Ground Map Fusion project type. Point it to the parent folder containing both the Drone/ and Lixel/ subfolders. LixelStudio will read the project structure and configure the merge parameters. Verify the coordinate system and run processing.
Processing: LCC Studio vs LixelStudio in Detail
Both pipelines take the same field data and produce different outputs. The decision between them is a deliverable decision, not a quality decision. A Gaussian Splat from LCC Studio is not better or worse than a point cloud from LixelStudio. They serve different purposes.
You can run both pipelines on the same field data. If you collect the ground scan with an L2 Pro and organize the drone data with both the LCC Studio folder structure and the LixelStudio Drone/Lixel structure, you can process in both applications and deliver either a Gaussian Splat or a point cloud (or both) from a single field session. This requires slightly more organized file management but no additional field time. Do not run both processing jobs simultaneously on the same machine. Process one pipeline to completion before starting the other, or use two separate workstations.
©2026 Alpine Reality Capture LLC • XGRIDS Pro Guide™ • Site Disclaimer