Neural Implicit Surface Reconstruction using Imaging Sonar
This addresses the problem of underwater or low-visibility environment mapping for robotics or marine applications, offering an incremental improvement over existing sonar-based methods.
The paper tackles dense 3D reconstruction of objects using imaging sonar by representing geometry as a neural implicit function and using a differentiable volumetric renderer for acoustic wave propagation, achieving higher-fidelity surface geometry from multi-view images compared to previous techniques with reduced memory overhead.
We present a technique for dense 3D reconstruction of objects using an imaging sonar, also known as forward-looking sonar (FLS). Compared to previous methods that model the scene geometry as point clouds or volumetric grids, we represent the geometry as a neural implicit function. Additionally, given such a representation, we use a differentiable volumetric renderer that models the propagation of acoustic waves to synthesize imaging sonar measurements. We perform experiments on real and synthetic datasets and show that our algorithm reconstructs high-fidelity surface geometry from multi-view FLS images at much higher quality than was possible with previous techniques and without suffering from their associated memory overhead.