Probabilistic Volumetric Fusion for Dense Monocular SLAM

TL;DR

This paper introduces a probabilistic volumetric fusion method that enhances dense monocular SLAM by using depth uncertainty derived from bundle adjustment, resulting in more accurate and artifact-free 3D reconstructions in real-time.

Contribution

It presents a novel way to estimate depth uncertainty directly from SLAM bundle adjustment, improving volumetric fusion quality over previous methods.

Findings

Achieves 92% better accuracy than direct depth fusion from monocular SLAM.

Up to 90% improvement over the best existing approaches.

Produces dense, accurate, and artifact-free 3D meshes in real-time.

Abstract

We present a novel method to reconstruct 3D scenes from images by leveraging deep dense monocular SLAM and fast uncertainty propagation. The proposed approach is able to 3D reconstruct scenes densely, accurately, and in real-time while being robust to extremely noisy depth estimates coming from dense monocular SLAM. Differently from previous approaches, that either use ad-hoc depth filters, or that estimate the depth uncertainty from RGB-D cameras' sensor models, our probabilistic depth uncertainty derives directly from the information matrix of the underlying bundle adjustment problem in SLAM. We show that the resulting depth uncertainty provides an excellent signal to weight the depth-maps for volumetric fusion. Without our depth uncertainty, the resulting mesh is noisy and with artifacts, while our approach generates an accurate 3D mesh with significantly fewer artifacts. We provide results on the challenging Euroc dataset, and show that our approach achieves 92% better accuracy than directly fusing depths from monocular SLAM, and up to 90% improvements compared to the best competing approach.