Abstract:Conventional geometry-based SLAM systems lack dense 3D reconstruction capabilities since their data association usually relies on feature correspondences. Additionally, learning-based SLAM systems often fall short in terms of real-time performance and accuracy. Balancing real-time performance with dense 3D reconstruction capabilities is a challenging problem. In this paper, we propose a real-time RGB-D SLAM system that incorporates a novel view synthesis technique, 3D Gaussian Splatting, for 3D scene representation and pose estimation. This technique leverages the real-time rendering performance of 3D Gaussian Splatting with rasterization and allows for differentiable optimization in real time through CUDA implementation. We also enable mesh reconstruction from 3D Gaussians for explicit dense 3D reconstruction. To estimate accurate camera poses, we utilize a rotation-translation decoupled strategy with inverse optimization. This involves iteratively updating both in several iterations through gradient-based optimization. This process includes differentiably rendering RGB, depth, and silhouette maps and updating the camera parameters to minimize a combined loss of photometric loss, depth geometry loss, and visibility loss, given the existing 3D Gaussian map. However, 3D Gaussian Splatting (3DGS) struggles to accurately represent surfaces due to the multi-view inconsistency of 3D Gaussians, which can lead to reduced accuracy in both camera pose estimation and scene reconstruction. To address this, we utilize depth priors as additional regularization to enforce geometric constraints, thereby improving the accuracy of both pose estimation and 3D reconstruction. We also provide extensive experimental results on public benchmark datasets to demonstrate the effectiveness of our proposed methods in terms of pose accuracy, geometric accuracy, and rendering performance.

Three-dimensional Dynamic Surface Alignment Based on Isometric Random Walk Graph

Maximum Spatial–temporal Isometric Cluster for Dynamic Surface Correspondence

A Submap Joining Algorithm for 3D Reconstruction Using an RGB-D Camera Based on Point and Plane Features

Probabilistic 3D ICP algorithm based on ORB feature

Bayesian mesh reconstruction from noisy point data

Face Alignment Based on 3D Face Shape Model and Markov Random Field.

Prior surface model guided dense reconstruction via graph-cuts

Robotic Trajectory Planning for Non-Destructive Testing Based on Surface 3D Point Cloud Data

Space-time 2D Gaussian Splatting for Accurate Surface Reconstruction under Complex Dynamic Scenes

A Stochastic Algorithm for 3D Scene Segmentation and Reconstruction

Triangular Mesh Generation for a Real-structure Scan Data in 3D Space

Surface Coverage Algorithm in Directional Sensor Networks for Three-Dimensional Complex Terrains

3D Dynamic Point Cloud Inpainting Via Temporal Consistency on Graphs

A New Approach to 3D Object Surface Reconstruction from Unorganized Points

A spatial overlay analysis method for three-dimensional vector polyhedrons

Surface Reconstruction for 3D Objects from Unorganized Points

A Novel Algorithm for Tracking Step-Like Edge Surfaces within 3D Images

Visual SLAM with 3D Gaussian Primitives and Depth Priors Enabling Novel View Synthesis

Robust 3d Surface Reconstruction In Real-Time With Localization Sensor

A Front Advancing Adaptive Triangular Mesh Dynamic Generation Algorithm and Its Application in 3D Geological Modeling

A Geometric Spread Approach of 3-D Reconstruction