Skip to content

Dense Reconstruction

Generates dense point clouds or 3D meshes from oriented photo sets.

Parameter Settings

General Settings

  • Use GPU: Enabled by default (recommended).

Densify Point Cloud

  • Resolution-Level: Reduces input photo resolution by a factor of N for faster processing.

  • Max-Resolution: Limits maximum processing resolution (combined with Resolution-Level).

  • Min-Resolution: Sets minimum resolution threshold.

Reconstruct Mesh

  • Close-Holes: 30 (default). Closes holes smaller than the specified diameter (in pixels).

  • Smooth: 2 (default). Controls mesh surface smoothness.

Notes

  • Default Behavior:

    • Generates dense point clouds only (mesh reconstruction is disabled).

    • In Photo Modeling Workflow, mesh reconstruction is automatically enabled.

  • Mask Integration:

    • Uses MASK (automatically generated during Mainbody Detection) to crop the dense point cloud/mesh.

Fine Registeration

Photo-to-Model Fine Registration

Iteratively optimizes the spatial alignment between photo set point clouds and the 3D model to enhance mapping accuracy.

Notes

Pre-Registration Preparation:

Use Segment to remove irrelevant point cloud regions.

Apply SOR Filter to eliminate outliers.

Model-to-Model Fine Registration

Aligns different LODs (Levels of Detail) of the same model in position and scale.

Workflow:

  • Roughly align models using Transform / Rotate tools.

  • Enable Fine Registration to refine alignment.

Notes

  • Scanned Models: Disable Adjust Scale to preserve original dimensions.

  • Symmetry Check: For symmetric objects (e.g., vases), manually verify alignment from multiple views.

Technical Terms

  • LOD (Level of Detail): Variants of a model with differing polygon counts.

  • SOR Filter: Statistical Outlier Removal filter.