Single View Seafloor Recovery from Imaging Sonar via Differentiable Rendering

2026-05-22

AI Summary

Recovering 3D seafloor topography from a single forward-looking sonar image is highly challenging due to vertical structure collapse, inherent ambiguities, and the reliance of existing methods on multi-view observations, multi-sensor fusion, or large annotated training datasets. This work proposes a training-free, physics-driven approach that, for the first time, introduces differentiable rendering to monocular sonar height estimation. By leveraging a differentiable sonar ray-tracing model and incorporating prior knowledge of seabed slope, the method directly optimizes an explicit elevation field to reproduce the observed sonar image. The reconstruction completes within 30 seconds, requires no training, and readily generalizes across diverse sensors and environments, effectively mitigating distributional shift. Experiments demonstrate that it outperforms supervised CNNs on out-of-distribution and complex synthetic terrains, while slightly underperforming only in in-distribution scenarios, thereby enabling rapid, training-free, and cross-environment seafloor reconstruction.

Abstract

Sonar is often the only modality suitable for high-resolution imaging underwater due to light attenuation and turbidity. Forward-looking imaging sonar provides measurements over range and horizontal angle but collapses vertical structure into a flat image, creating ambiguities that make 3D recovery challenging. A common use case for imaging sonar is underwater terrain mapping (bathymetry), yet current methods require many views, expensive multi-sensor setups, or significant training data, which limits use and adaptability to new environments. We present a training-free method that recovers bathymetry from a single sonar image in under 30 seconds via differentiable rendering, conditioned on a known seafloor tilt. To our knowledge, this is the first differentiable rendering approach for single-view height recovery in sonar. Our method implements differentiable sonar ray tracing and optimizes an explicit height field to reproduce the target image. On synthetic datasets, our approach outperforms a supervised CNN under distribution shift and remains close on rough terrain, while the CNN wins in-distribution. By modeling physically grounded priors of the sonar process, our method adapts across sensor configurations and environments without training data.

Problem

• single-view
• seafloor recovery
• imaging sonar
• bathymetry
• 3D reconstruction

Innovation

• differentiable rendering
• single-view bathymetry
• imaging sonar
• training-free reconstruction
• sonar ray tracing