SAR-NeRF: Neural Radiance Fields for Synthetic Aperture Radar Multi-View Representation

TL;DR

SAR-NeRF introduces a neural radiance field model tailored for SAR images, effectively capturing view-dependent features and enhancing multi-view representation and classification performance, especially in few-shot scenarios.

Contribution

This work develops a novel SAR-NeRF model that combines SAR imaging principles with neural networks to improve multi-view representation and generalization of SAR images.

Findings

SAR-NeRF accurately models SAR images as functions of attenuation and scattering in 3D space.

The model significantly improves SAR target classification accuracy in few-shot learning.

SAR-NeRF enhances multi-view generalization capabilities of SAR image representations.

Abstract

SAR images are highly sensitive to observation configurations, and they exhibit significant variations across different viewing angles, making it challenging to represent and learn their anisotropic features. As a result, deep learning methods often generalize poorly across different view angles. Inspired by the concept of neural radiance fields (NeRF), this study combines SAR imaging mechanisms with neural networks to propose a novel NeRF model for SAR image generation. Following the mapping and projection pinciples, a set of SAR images is modeled implicitly as a function of attenuation coefficients and scattering intensities in the 3D imaging space through a differentiable rendering equation. SAR-NeRF is then constructed to learn the distribution of attenuation coefficients and scattering intensities of voxels, where the vectorized form of 3D voxel SAR rendering equation and the sampling relationship between the 3D space voxels and the 2D view ray grids are analytically derived. Through quantitative experiments on various datasets, we thoroughly assess the multi-view representation and generalization capabilities of SAR-NeRF. Additionally, it is found that SAR-NeRF augumented dataset can significantly improve SAR target classification performance under few-shot learning setup, where a 10-type classification accuracy of 91.6\% can be achieved by using only 12 images per class.