PhysDNet: Physics-Guided Decomposition Network of Side-Scan Sonar Imagery

2025-11-24

AI Summary

Side-scan sonar (SSS) image intensity is jointly governed by seafloor reflectivity, terrain elevation, and acoustic path loss, resulting in strong viewpoint dependence and physical inconsistency—hindering downstream tasks such as image registration and shadow interpretation. To address this, we propose a physics-guided, self-supervised multi-branch neural network that decomposes SSS imagery into three physically interpretable components: reflectivity, terrain elevation, and propagation loss. Our method integrates the Lambertian reflectance model with deep learning, enforcing explicit physical constraints to enable end-to-end reconstruction without ground-truth supervision. Experiments demonstrate that the decomposition preserves geological structural coherence, generates illumination and attenuation fields consistent with underwater acoustics, and yields high-fidelity shadow maps. This significantly enhances analytical robustness and physical fidelity—enabling more reliable seabed characterization and geometric reasoning from SSS data.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Side-scan sonar (SSS) imagery is widely used for seafloor mapping and underwater remote sensing, yet the measured intensity is strongly influenced by seabed reflectivity, terrain elevation, and acoustic path loss. This entanglement makes the imagery highly view-dependent and reduces the robustness of downstream analysis. In this letter, we present PhysDNet, a physics-guided multi-branch network that decouples SSS images into three interpretable fields: seabed reflectivity, terrain elevation, and propagation loss. By embedding the Lambertian reflection model, PhysDNet reconstructs sonar intensity from these components, enabling self-supervised training without ground-truth annotations. Experiments show that the decomposed representations preserve stable geological structures, capture physically consistent illumination and attenuation, and produce reliable shadow maps. These findings demonstrate that physics-guided decomposition provides a stable and interpretable domain for SSS analysis, improving both physical consistency and downstream tasks such as registration and shadow interpretation.

Problem

• Decouples side-scan sonar images into three interpretable physical fields
• Resolves view-dependency issues caused by entangled acoustic factors
• Enables self-supervised training without ground-truth annotations

Innovation

• Physics-guided network decouples sonar imagery components
• Self-supervised training using Lambertian reflection model
• Decomposed representations enable stable geological structure preservation