SonarSplat: Novel View Synthesis of Imaging Sonar via Gaussian Splatting

2025-03-31

AI Summary

This work addresses two key challenges in imaging sonar novel-view synthesis: inaccurate acoustic scattering modeling and azimuthal smearing artifacts. To this end, we propose the first 3D Gaussian-based sonar imaging framework. Methodologically: (1) underwater scenes are modeled as differentiable 3D Gaussian primitives endowed with spatially varying acoustic reflectivity and saturation characteristics; (2) Gaussian splatting is introduced for sonar for the first time, coupled with a physics-informed rasterization operator tailored to sonar’s wave propagation and beamforming properties; (3) azimuthal smearing is explicitly parameterized, enabling joint smearing suppression and geometric reconstruction. Evaluated on real-world underwater robot datasets, our method achieves a 2.5 dB PSNR improvement over state-of-the-art approaches. It simultaneously delivers high-fidelity novel-view synthesis, effective smearing removal, and dense 3D geometric reconstruction.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In this paper, we present SonarSplat, a novel Gaussian 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 learned 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 (+2.5 dB PSNR). We also demonstrate that SonarSplat can be leveraged for azimuth streak removal and 3D scene reconstruction.

Problem

• Novel view synthesis for imaging sonar using Gaussian splatting
• Modeling acoustic streaking phenomena in sonar images
• Improving image synthesis and 3D reconstruction accuracy

Innovation

• Gaussian splatting for sonar view synthesis
• Acoustic reflectance in 3D Gaussians
• Novel azimuth streaking modeling