Source scan
25M+ points
The original Luminy LiDAR tile before filtering
VR / Mesh Processing / 3D Reconstruction / Computer Graphics2025Research Project
Interactive Geometry Processing in VR
A Unity VR experience that turns LiDAR scans into lightweight 3D buildings and lets players cut virtual objects naturally with a sword.
UnityVRLiDAROpen3DMesh Processing
Results
Visual metrics
Building data
70K-80K points
Typical isolated point cloud used for one building
Unity-ready model
2K-5K faces
Simplified range used to keep the scene responsive
Cut repair time
~3 ms
Fast radial method on a boundary of about 50 points
Technical snapshot
Quick facts
Goal
From real-world scans to interactive VR geometry
Role
Geometry processing and Unity VR implementation
Duration
8-week Master GIG project, 2026
Stack
Python, Open3D, CloudCompare, Unity
Dataset
IGN LiDAR HD data from the Luminy area
Output
Reconstructed building and VR slicing demo
Status
Research Project
Case study
Overview
Problem
LiDAR data contains millions of separate points rather than ready-to-use surfaces, so it is too heavy and incomplete for direct use in Unity. At the same time, many games only play a prepared cutting animation, meaning an object always breaks in the same way regardless of how the player swings the tool.
Goal
Turn large sets of scanned points into usable 3D models, then create a convincing VR cutting interaction that follows the position and angle chosen by the player.
Method
The team filtered IGN LiDAR data to isolate buildings, joined nearby points into surfaces, repaired missing areas and simplified each model for real-time display. In Unity, a dynamic cutting system divides an object along the player's movement and rebuilds the newly exposed interior with colored layers.
Result
The project produced a recognizable low-poly building from real Luminy scan data and an interactive VR demonstration where objects split into two parts without noticeable delay. A faster method for rebuilding the cut interior reduced a representative calculation from 2.7 seconds to about 3 milliseconds.
Technical foundation
Technology stack
Data
IGN LiDAR HDCloudCompare
Geometry
PythonOpen3DBall Pivoting
Runtime
UnityC#VR
Interaction
Dynamic Mesh SlicerVolumetric Mesh
Evolution
Research timeline
2026
LiDAR reconstruction
Filtered the Luminy dataset and turned building point clouds into repaired, simplified surfaces.
2026
Dynamic mesh cutting
Connected the player's sword movement to a cut that divides the actual geometry in real time.
2026
VR demonstration
Combined the reconstructed building and slicing system in an explorable Unity environment.
Comparison
From real building to reconstructed geometry
The real Luminy building used as a visual reference and the lightweight 3D model reconstructed from its LiDAR scan for use in Unity.
Real-world building
Reconstructed 3D model
Process
Project pipeline
01
Collect LiDAR data
Start with a large IGN scan made of millions of points representing terrain, trees and buildings.
Input
02
Isolate each building
Filter the scan and separate useful building points from the surrounding environment.
Preparation
03
Build and repair surfaces
Connect the points into visible surfaces, close small gaps and correct incorrectly facing areas.
Reconstruction
04
Prepare for Unity
Reduce the number of faces so the reconstructed building remains recognizable but runs smoothly in VR.
Optimization
05
Cut objects in VR
Use the sword movement to define the real position and angle of each cut instead of playing a fixed animation.
Interaction
06
Rebuild the interior
Close the cut immediately and add colored inner layers before displaying the two resulting objects.
Output
01
Collect LiDAR data
Start with a large IGN scan made of millions of points representing terrain, trees and buildings.
Input
02
Isolate each building
Filter the scan and separate useful building points from the surrounding environment.
Preparation
03
Build and repair surfaces
Connect the points into visible surfaces, close small gaps and correct incorrectly facing areas.
Reconstruction
04
Prepare for Unity
Reduce the number of faces so the reconstructed building remains recognizable but runs smoothly in VR.
Optimization
05
Cut objects in VR
Use the sword movement to define the real position and angle of each cut instead of playing a fixed animation.
Interaction
06
Rebuild the interior
Close the cut immediately and add colored inner layers before displaying the two resulting objects.
Output
Project evidence
Media gallery
video
Interactive VR Geometry Demonstration
Project demonstration showing the reconstructed environment and real-time geometry interactions in Unity VR.
image
Volumetric Mesh in Unity
The reconstructed interior layers shown after an object is cut in the VR experience.
3D model
Reconstructed Luminy Building
Loading 3D viewer
Model slot for a Unity-ready building reconstructed from IGN LiDAR data.
Implementation detail
Technical notes
- The reconstruction starts from classified LiDAR data, which makes it easier to separate buildings from terrain and vegetation.
- The generated building is simplified before import because raw scan geometry is too heavy for a responsive VR scene.
- Each sword movement creates a different cut based on the player's chosen position and angle.
- The new interior is divided into colored layers to make the volume of the cut object easy to understand.
- The fast interior-filling method currently works best on closed, convex objects.
- Future work includes more reliable surface orientation and support for more complex, non-convex shapes.
Knowledge base
References
website
IGN LiDAR HD
The open geographic dataset used to reconstruct buildings from the Luminy area.
documentation
Open3D
The 3D data library used for point-cloud processing, surface reconstruction and mesh simplification.
documentation
Unity XR Documentation
Unity guidance for building and testing immersive applications.
References
Resources
Keep exploring
