Automated Surface Skinning in BRL-CAD

[brlcad]

Automated Surface Skinning: Exterior Mesh Extraction and Simplification for BRL-CAD Models

Outline

BRL-CAD models often consist of complex, nested CSG hierarchies that are computationally expensive to render or export for external applications. "Skinning" refers to the process of generating a simplified, watertight exterior polygonal mesh that represents the outer boundary of a model while discarding internal geometry. This project aims to implement a tool—either through a new algorithm or the integration of license-compatible libraries—to automate the creation of these exterior "shells."

Details

The project will focus on transforming internal CSG representations into optimized boundary representations (B-Reps). Key technical components include:

• Algorithm Selection: Evaluating and implementing algorithms such as Voxel-based Shrink-wrapping, Alpha Shapes, or Visible Surface Manifold extraction.
• Internal Geometry Pruning: Developing logic to identify and ignore "hidden" internal components that do not contribute to the exterior silhouette.
• Mesh Simplification: Once the skin is generated, implementing decimation techniques (like Quadratic Error Metrics) to reduce polygon count while preserving visual fidelity.
• Integration: Creating a standalone CLI tool or a dedicated command within MGED/Archer. If a semi-automated route is chosen, this would involve a GUI allowing users to "paint" or select focus areas for skinning.

Expected Outcome

• A functional BRL-CAD utility (e.g., g-skin or new GUI) that accepts a .g database and outputs a simplified, manifold .g or .gltf file.
• A robust method for handling "gaps" or "leaks" in the mesh to ensure the output is watertight.
• Documentation covering the algorithm's constraints and a performance comparison against raw mesh exports.

Future Possibilities

• LOD Generation: Automatically creating various "Levels of Detail" (LOD) for real-time visualization.
• Thermal/CFD Pre-processing: Enhancing the tool to produce meshes specifically optimized for simulation solvers.
• GUI Plugin: A full integration into the MGED or Archer interface with real-time "shrink-wrap" previews.

Project Properties

Skills

• Languages: C/C++
• Mathematics: Computational geometry and mesh topology optional but helpful.
• APIs: Familiarity with BRL-CAD's librt and libwdb libraries and related conversion commands optional to know in advance but necessary for implementation.

Difficulty

Medium or Hard depending on whether you plan to implement your own algorithm or glue

Size

Long (350h)

Additional Information

• Potential mentor(s): Sean Morrison
  Erik Greenwald
• Organization website: https://brlcad.org
• Communication channels: https://brlcad.zulipchat.com

[Mahek-01001]

Hi @brlcad and mentors,

I’ve been digging into the Automated Skinning project (Issue #106) and wanted to share a functional prototype I put together to test out the voxel-based shrink-wrapping approach.

I used a high-detail Bugatti model as a "torture test" to see if I could effectively "hollow out" a complex assembly. I'm pretty happy with the results—the algorithm managed to prune all the internal clutter (engine, seats, etc.) while keeping a perfectly watertight exterior shell.

A few technical notes on the current pipeline:

Internal Pruning: I’m filling the voxel grid to make sure everything inside the silhouette is ignored.

Watertight Output: Using Marching Cubes ensures the resulting mesh is a closed manifold, which is critical for BRL-CAD's solid modeling.

Simplification: I also added a QEM decimation step. It managed to knock the face count down by about 95% without losing the car's shape or sharp curves.

You can see the "Ghost Overlay" (the skin fitting over the original geometry) and the full code in this repo:
🔗 https://github.com/Mahek-01001/brlcad-skinning-prototype/tree/main

I’m currently spending my time in librt and libwdb to map out the best way to port this logic into a native C++ utility. I'd love to get some early feedback on this workflow, and I'm looking forward to discussing this more with the team!