Neural Stochastic Screened Poisson Reconstruction
This work addresses surface reconstruction uncertainty for 3D scanning applications, but it appears incremental as it builds on existing Poisson-based methods with neural enhancements.
The paper tackles the problem of reconstructing surfaces from point clouds by using a neural network to quantify reconstruction uncertainty under a Poisson smoothness prior, enabling integration into a full 3D scanning pipeline for iterative updates.
Reconstructing a surface from a point cloud is an underdetermined problem. We use a neural network to study and quantify this reconstruction uncertainty under a Poisson smoothness prior. Our algorithm addresses the main limitations of existing work and can be fully integrated into the 3D scanning pipeline, from obtaining an initial reconstruction to deciding on the next best sensor position and updating the reconstruction upon capturing more data.