5. 3DSC for RealityCapture

3DSC for RealityCapture is an integration module within the 3D Survey Collection add-on that enables bi-directional workflow between Blender and Capturing Reality’s RealityCapture software. The module provides tools for exporting models, managing LOD (Level of Detail) structures, importing reconstruction regions, and automating texture generation.

5.1. Overview 

The RealityCapture integration in 3DSC allows users to:

Export 3D models directly to RealityCapture with configurable quality parameters
Import OBJ files processed by RealityCapture back into Blender
Automatically organize imported LOD structures into Blender collections
Correct naming conventions for mesh, materials, and objects
Import RealityCapture reconstruction regions (.rcbox files) as 3D geometry
Manage texturing workflows with precise texel size control

Platform Note

The RealityCapture Integration panel is only visible on Windows systems, as RealityCapture is a Windows-exclusive application.

5.2. Installation & Requirements 

Requirements:

Windows 10/11 (64-bit)
Capturing Reality RealityCapture (any version with CLI support)
Blender 4.2 or later
3DSC Add-on installed and enabled

Configuration:

Open Blender and navigate to the 3DSC sidebar panel
Expand the Reality Capture Integration section
Set the RC Executable Path to your RealityCapture installation:

Default: C:\Program Files\Capturing Reality\RealityCapture\RealityCapture.exe
Set the Project Path to your .rcproj file
Define an Exchange Folder for file transfer between applications

5.3. Parameters Reference 

The following parameters control the RealityCapture integration:

Parameter	Default	Description
RC Executable Path	(system default)	Path to RealityCapture.exe
Project Path	(empty)	Path to .rcproj project file
Exchange Folder	(empty)	Folder for import/export operations
Max Resolution	4096	Maximum texture resolution (1024-16384)
Detail Levels	5	Number of LOD levels to generate (1-10)
Texel Size	0.01	Target texel size in meters (0.001-1.0)
Texture Resolution	2048	Texture resolution for export (256-16384)

5.4. Tool Reference 

5.4.1. Export Tools 

Export to RC

Exports the current Blender scene data to RealityCapture with configured parameters.

Operator: object.export_rc

Workflow:

Configure export parameters (Max Resolution, Detail Levels, Texel Size)
Click Export to RC
Files are saved to the Exchange Folder
RealityCapture processes the export with specified settings

Export Cleaned OBJ

Exports a cleaned mesh as OBJ to the Exchange Folder.

Operator: object.export_cleaned_obj

Output: cleaned.obj in Exchange Folder

Export LOD to Exchange Folder

Exports Level of Detail models from RealityCapture.

Operator: object.export_lod

Features:

Queries model name from RC project
Selects specified model
Exports LOD structure to Exchange Folder

5.4.2. Import Tools 

Import OBJ

Imports all OBJ files from the Exchange Folder into Blender.

Operator: object.import_obj

Features:

Batch import of multiple OBJ files
Automatic detection of files in Exchange Folder

Import Reconstruction Region

Imports RealityCapture .rcbox files as 3D geometry in Blender.

Operator: import.reconstruction_region

File Format: XML-based ReconstructionRegion files

Parameters:

Apply Shift: When enabled, applies global coordinate shift to the imported geometry

Generated Geometry:

Creates a 3D cube representing the reconstruction region
Applies correct dimensions (width, height, depth)
Sets rotation from yaw/pitch/roll values
Optionally applies coordinate transformation

5.4.3. Texturing Tools 

Texture in RC

Launches the texturing process in RealityCapture with specified parameters.

Operator: object.texture_rc

Parameters Applied:

Texel Size: Controls texture detail level
Texture Resolution: Maximum texture dimensions

RC Commands: Sets parameters and executes -texture all

5.4.4. LOD Management 

Organize LODs to Collections

Automatically organizes imported LOD objects into Blender collections based on their naming suffix.

Operator: organize.lods_to_collections

Pattern Detection: Identifies LOD0, LOD1, LOD2, etc. suffixes

Result: Creates collections named LOD0, LOD1, LOD2 and moves objects accordingly

Correct LOD Names (Mesh & Materials)

Standardizes naming conventions for objects, meshes, and materials imported from RealityCapture.

Operator: correct.rcnames

Naming Patterns Supported:

Old format: base_LODx_number converted to base_number_LODx
New format: base_number_LODx (already correct, no change)
Generic fallback for unusual cases

Scope:

Renames selected objects
Updates associated mesh data names
Corrects material names

Best Practice

Always run Correct LOD Names before Organize LODs to Collections to ensure proper detection and organization.

5.5. Complete Workflow Example 

This example demonstrates a typical workflow using both Blender and RealityCapture.

5.5.1. Phase 1: Setup 

Open your RealityCapture project with processed photogrammetric data
In Blender, configure 3DSC RealityCapture settings:
- Set RC Executable Path
- Set Project Path to your .rcproj file
- Define Exchange Folder

5.5.2. Phase 2: Export from RealityCapture 

In RealityCapture, complete your mesh reconstruction
In Blender, use Export LOD to export LOD structure
RealityCapture generates multi-level detail models to Exchange Folder

5.5.3. Phase 3: Import and Organize in Blender 

Click Import OBJ to batch import all models
Select all imported objects
Run Correct LOD Names to standardize naming
Run Organize LODs to Collections to create LOD hierarchy

5.6. Multi-Sensor Alignment Tutorial 

This section provides detailed guidance on aligning data from multiple sensors (laser scanner, drone, different camera lenses) in RealityCapture. This workflow is particularly useful for complex archaeological or architectural documentation projects.

5.6.1. Overview of Multi-Sensor Approach 

When working with multiple acquisition devices, each sensor produces data with different characteristics. Proper alignment requires careful project organization and parameter tuning.

Typical Sensor Combinations:

Laser Scanner data
Drone imagery (aerial)
Ground-based photography (24mm lens)
Ground-based photography (35mm lens)

5.6.2. Project Structure 

Create separate projects for each sensor combination:

Project	Content	Notes
1	Laser Scanner + Drone only	Export component after alignment
2	Laser Scanner + 24mm only	Export component after alignment
3	Laser Scanner + 35mm only	Export component after alignment
4	Laser + Drone + 24mm + 35mm	Master project - Import and merge

5.6.3. Control Points Setup 

Important

Use Control Points, NOT Ground Control Points (GCPs), to verify alignment between images and laser data.

Best Practices:

Place control points at features visible in both laser and photographic data
Use points in common between sensor types to verify alignment quality
Consider using automatic target detection or scalebars for improved results

5.6.4. Workflow Phase 1: Individual Sensor Projects (1-3)

For each single-sensor project (Projects 1, 2, and 3):

Import Data

Import laser scanner data and the corresponding image set (drone, 24mm, or 35mm)
Configure Alignment Settings

Apply the optimized alignment parameters (see next section)
Run Alignment

Execute the alignment process
Place Control Points

Add control points at common features between laser and images
Verify Alignment Quality

Check that control points show consistent positions
Export Component

Use Workflow > Export Component to save the aligned dataset

5.6.5. Workflow Phase 2: Master Project (4)

In the master project, combine all sensor data:

Import Component

Use Workflow > Import Component for each exported component (one at a time)
Merge Component

When prompted “with images?”, select Yes
Sensor Lock Option

You can choose to NOT move sensors during merge. This prevents RealityCapture from forcing alignment and should be disabled in the alignment menu if needed.
Rebuild Sparse Cloud

Regenerate the sparse point cloud while keeping sensors fixed (alignment)
Image Usage Options

You can disable images from being used for geometry while keeping them aligned for texture purposes

Technical Note

When merging components, RealityCapture may attempt to force alignment between sensors. If you have already verified alignment quality with control points, you can disable automatic sensor movement in the alignment settings.

5.6.6. Alignment Settings Reference 

Use these optimized settings for multi-sensor alignment:

Parameter	Value	Notes
Max feature reprojection error	3.0 - 4.0	Warning: 2.0 eliminates too many good tie points!
Feature detection quality	High	Keep default setting
Max features per mpx	30,000 - 50,000	Higher values improve matching
Max features per image	100,000 - 150,000	Increase for complex scenes
Image overlap	High	Essential for multi-sensor work
Image downscale factor	1	Full resolution (optimal)

Why these values?

Reprojection error 3.0-4.0: A value of 2.0 is too aggressive and removes valid tie points, especially when matching between different sensor types
High feature counts: Multi-sensor alignment benefits from more features to find correspondences between different image qualities
No downscaling: Full resolution ensures maximum feature detection accuracy

5.6.7. Tips for Multi-Sensor Success 

1. Automatic Target Recognition

Use coded targets or scalebars that can be automatically detected by RealityCapture. This significantly improves:

Initial alignment accuracy
Scale consistency between sensors
Verification of final results

2. Selective Geometry Usage

You can configure images to:

Contribute to geometry: Images are used for mesh reconstruction
Texture only: Images are used only for texturing, not geometry

This is useful when one sensor (e.g., laser) provides better geometry while another (e.g., 24mm photos) provides better texture.

3. Component Export/Import Strategy

Export each sensor combination as a separate component. This allows:

Independent quality verification
Rollback capability if merge fails
Modular project management

4. Alignment Locking

After achieving good alignment, you can lock sensors to prevent unintended changes during subsequent operations:

Prevents accidental misalignment during merge
Maintains verified control point accuracy
Allows texture-only operations without geometry changes

5.7. Troubleshooting 

5.7.1. Common Issues and Solutions 

Issue: Panel not visible in Blender

Cause: Running on macOS or Linux

Solution: RealityCapture integration requires Windows. Use the CLI interface if cross-platform operation is needed.

Issue: “RC Executable not found”

Cause: Incorrect path configuration

Solution:

Verify the path to RealityCapture.exe
Check that RealityCapture is properly installed
Try running RealityCapture manually to confirm installation

Issue: Import fails with empty result

Cause: No OBJ files in Exchange Folder

Solution:

Verify files exist in the Exchange Folder
Check file permissions
Ensure export from RealityCapture completed successfully

Issue: LOD organization creates wrong collections

Cause: Non-standard naming in imported files

Solution:

Run Correct LOD Names first
Verify naming pattern matches *_LOD# format
Manually rename problematic objects if needed

Issue: Reconstruction Region imports at wrong location

Cause: Coordinate system mismatch

Solution:

Check the Apply Shift option
Verify the .rcbox file contains correct coordinates
Compare with known reference points in your scene

5.7.2. ReconstructionRegion File Format 

The .rcbox file is an XML format containing:

<ReconstructionRegion>
  <globalCoordinateSystem>...</globalCoordinateSystem>
  <globalCoordinateSystemWkt>...</globalCoordinateSystemWkt>
  <globalCoordinateSystemName>...</globalCoordinateSystemName>
  <isGeoreferenced>true/false</isGeoreferenced>
  <isLatLon>true/false</isLatLon>
  <yawPitchRoll>Y P R</yawPitchRoll>
  <widthHeightDepth>W H D</widthHeightDepth>
  <Header version="X" magic="Y"/>
  <CentreEuclid>X Y Z</CentreEuclid>
  <Residual R="..." t="..." s="..." ownerId="..."/>
</ReconstructionRegion>

Key Elements:

yawPitchRoll: Rotation angles for the region box
widthHeightDepth: Dimensions of the reconstruction volume
CentreEuclid: Center coordinates in Euclidean space
isGeoreferenced: Indicates if coordinates are georeferenced

5.8. Credits and License 

Author:: Emanuel Demetrescu CNR-ISPC (Consiglio Nazionale delle Ricerche - Istituto di Scienze del Patrimonio Culturale) Rome, Italy
License:: GNU General Public License v3 (GPL-3)
Part of:: Extended Matrix Framework https://www.extendedmatrix.org