Research Analyzer
← Back ICRA 2026

SonarSplat: Novel View Synthesis of Imaging Sonar Via Gaussian Splatting

Advaith Venkatramanan Sethuraman, Max Rucker, Onur Bagoren, Pou-Chun Kung, Nibarkavi Naresh Babu Amutha, Katherine Skinner

PDF

AI summary

Key figure (auto-extracted from paper)
SonarSplat achieves high-fidelity, real-time sonar image synthesis and accurate 3D reconstruction by adapting 3D Gaussian splatting to model acoustic reflectance and azimuth streaking.
Imaging sonar Gaussian splatting Novel view synthesis 3D reconstruction Underwater robotics Azimuth streaking

Problem

Prior radiance field methods for sonar are either computationally prohibitive for real-time use or fail to accurately capture sonar-specific artifacts like azimuth streaking and elevation ambiguity, hindering reliable underwater perception.

Approach

The authors adapt 3D Gaussian splatting for imaging sonar by representing scenes as Gaussians with acoustic reflectance and streak probabilities, introducing a range/azimuth rasterization model, an adaptive gain module for streak effects, and a novel densification strategy to resolve elevation ambiguity.

Key results

  • +3.2 dB PSNR improvement over state-of-the-art novel view synthesis
  • 77% lower Chamfer Distance for more accurate 3D reconstruction
  • Effective removal of azimuth streaking artifacts from rendered sonar images
  • Introduction of Elevation Sampling Densification Strategy to resolve elevation ambiguity

Why it matters

Provides underwater roboticists with a fast, accurate tool for sonar-based mapping, inspection, and data augmentation that overcomes the speed and fidelity limits of prior neural rendering methods.

Abstract

In this paper, we present SonarSplat, a novel Gaus- sian splatting framework for imaging sonar that demonstrates realistic novel view synthesis and models acoustic streaking phenomena. Our method represents the scene as a set of 3D Gaussians with acoustic reflectance and saturation properties. We develop a novel method to efficiently rasterize Gaussians to produce a range/azimuth image that is faithful to the acoustic image formation model of imaging sonar. In particular, we develop a novel approach to model azimuth streaking in a Gaussian splatting framework. We evaluate SonarSplat using real-world datasets of sonar images collected from an underwater robotic platform in a controlled test tank and in a real-world river environment. Compared to the state-of-the-art, SonarSplat offers improved image synthesis capabilities (+3.2 dB PSNR) and more accurate 3D reconstruction (77% lower Chamfer Distance). We also demonstrate that SonarSplat can be leveraged for azimuth streak removal.

Index terms

Marine Robotics Mapping Deep Learning for Visual Perception

Related papers