SurfelMeshing: Online Surfel-Based Mesh Reconstruction.

Area of Science:

• Computer Vision
• 3D Reconstruction
• Computer Graphics

Background:

• Real-time 3D mesh reconstruction from RGB-D video is crucial for applications like augmented reality and robotics.
• Existing methods often rely on volumetric fusion, which can struggle with thin objects and varying resolutions.

Purpose of the Study:

• To develop a novel surfel-based approach for dynamic mesh reconstruction from live RGB-D video.
• To enable real-time adaptation to scene changes and loop closures during Simultaneous Localization and Mapping (SLAM).

Main Methods:

• A dense surfel cloud representation is used instead of volumetric fusion.
• Asynchronous retriangulation of smoothed surfels reconstructs the surface mesh.
• The surfel cloud is deformed and remeshed to adapt to scene changes and loop closures.

Main Results:

• The surfel-based method maintains a dense surface representation adaptable to loop closures.
• It supports varying scan resolutions and reconstructs colors at the input camera's resolution.
• Thin objects are reconstructed effectively, outperforming volumetric approaches.

Conclusions:

• The proposed surfel-based approach offers competitive performance with state-of-the-art methods for mesh reconstruction.
• It provides advantages in handling dynamic scenes, varying resolutions, and thin object reconstruction.
• The algorithm is available as open-source software for further research and development.