EUVIMEDEUVIMED entdecken
Crossref

VIGS SLAM: IMU-Based Large-Scale RGB-D 3-D Gaussian Splatting SLAM

Gyuhyeon Pak, Euntai Kim

IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 5010510-5010510

DOI: 10.1109/tim.2026.3693421 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Gyuhyeon Pak und 1 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Gefördert von: Korea Evaluation Institute of Industrial Technology (KEIT) through Korean Government (MOTIE).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recently, map representations based on radiance fields, such as 3-D Gaussian splatting (3DGS) and neural radiance fields (NeRFs), which is excellent for realistic depiction, have attracted considerable attention, leading to attempts to combine them with simultaneous localization and mapping (SLAM). However, existing 3DGS-based frameworks are predominantly limited to room-scale environments due to their heavy reliance on dense photometric tracking, which is sensitive to data sparsity and computationally expensive. To overcome these limitations, we propose VIGS SLAM, a tightly coupled visual–inertial 3DGS SLAM framework designed for large-scale indoor scenarios. Our system employs a geometry-centric frontend based on generalized-ICP (GICP) that utilizes the probabilistic framework of Gaussians for robust tracking, significantly avoiding the computational overhead inherent in dense photometric tracking. To ensure global consistency in expansive regions, we introduce a tightly coupled factor graph backend that jointly minimizes IMU preintegration residuals and GICP-based pose residuals. Furthermore, a gravity-aligned initialization is incorporated to calculate the initial orientation, providing a reliable position that facilitates the observability of metric and promotes the rapid convergence of state variables, such as velocity and IMU biases, during subsequent optimization. This proposal not only enhances the performance of Gaussian splatting SLAM beyond room-scale scenarios but also achieves SLAM performance comparable with state-of-the-art methods in large-scale indoor environments. The source code is available at: https://github.com/ghPak1002/VIGS_SLAM

Förderung

  • Korea Evaluation Institute of Industrial Technology (KEIT) through Korean Government (MOTIE)

Zitieren

APA

Pak, G., & Kim, E. (2026). VIGS SLAM: IMU-Based Large-Scale RGB-D 3-D Gaussian Splatting SLAM. IEEE Transactions on Instrumentation and Measurement, 75, 5010510-5010510. https://doi.org/10.1109/tim.2026.3693421

Vancouver

Pak G, Kim E. VIGS SLAM: IMU-Based Large-Scale RGB-D 3-D Gaussian Splatting SLAM. IEEE Transactions on Instrumentation and Measurement. 2026;75:5010510-5010510. doi:10.1109/tim.2026.3693421

BibTeX
@article{Pak2026,
  author = {Pak, Gyuhyeon and Kim, Euntai},
  title = {VIGS SLAM: IMU-Based Large-Scale RGB-D 3-D Gaussian Splatting SLAM},
  journal = {IEEE Transactions on Instrumentation and Measurement},
  year = {2026},
  volume = {75},
  pages = {5010510--5010510},
  doi = {10.1109/tim.2026.3693421}
}
Kostenlos starten →