Technology Category
Reasoning under Uncertainty: Relational Probabilistic ModelsIntelligent Robots: State EstimationComputer Vision: 3D Computer Vision
BayesFusion-SDF: Probabilistic Signed Distance Fusion with View Planning on CPU
🤖 AI Summary
This work proposes the first CPU-based Bayesian probabilistic signed distance function (SDF) fusion framework, addressing key limitations of traditional TSDF methods—which rely on heuristic weighting and struggle to model geometric uncertainty—and neural implicit approaches, which are computationally expensive and lack interpretability. The method represents geometry as a sparse Gaussian random field and fuses depth observations through a heteroscedastic Bayesian formulation. To efficiently compute the posterior distribution, it introduces an adaptive narrow-band domain, a preconditioned conjugate gradient solver, and a stochastic diagonal estimator. Evaluated on CO3D and ablation studies, the approach achieves higher geometric accuracy than classical TSDF while providing reliable uncertainty estimates, enabling robust surface extraction and active viewpoint planning. This framework strikes a favorable balance among computational efficiency, reconstruction accuracy, and decision transparency.
Technology Category
Application Category
📝 Abstract
Key part of robotics, augmented reality, and digital inspection is dense 3D reconstruction from depth observations. Traditional volumetric fusion techniques, including truncated signed distance functions (TSDF), enable efficient and deterministic geometry reconstruction; however, they depend on heuristic weighting and fail to transparently convey uncertainty in a systematic way. Recent neural implicit methods, on the other hand, get very high fidelity but usually need a lot of GPU power for optimization and aren't very easy to understand for making decisions later on. This work presents BayesFusion-SDF, a CPU-centric probabilistic signed distance fusion framework that conceptualizes geometry as a sparse Gaussian random field with a defined posterior distribution over voxel distances. First, a rough TSDF reconstruction is used to create an adaptive narrow-band domain. Then, depth observations are combined using a heteroscedastic Bayesian formulation that is solved using sparse linear algebra and preconditioned conjugate gradients. Randomized diagonal estimators are a quick way to get an idea of posterior uncertainty. This makes it possible to extract surfaces and plan the next best view while taking into account uncertainty. Tests on a controlled ablation scene and a CO3D object sequence show that the new method is more accurate geometrically than TSDF baselines and gives useful estimates of uncertainty for active sensing. The proposed formulation provides a clear and easy-to-use alternative to GPU-heavy neural reconstruction methods while still being able to be understood in a probabilistic way and acting in a predictable way. GitHub: https://mazumdarsoumya.github.io/BayesFusionSDF
Problem
Research questions and friction points this paper is trying to address.
3D reconstruction
uncertainty quantification
signed distance function
probabilistic fusion
active sensing
Innovation
Methods, ideas, or system contributions that make the work stand out.
probabilistic reconstruction
signed distance function
Bayesian fusion
uncertainty quantification
active view planning
🔎 Similar Papers
No similar papers found.
