A robust method for high-resolution C/S-band SAR image matching via pseudo-modality generation and coupled spectrum-spatial learning

Mingyang Song

doi:10.1016/j.rse.2026.115612

A robust method for high-resolution C/S-band SAR image matching via pseudo-modality generation and coupled spectrum-spatial learning

2026-08-09

0

PRE

AI

M

Mingyang Song

H

Hong Zhang *

Y

Yukun Fan

L

Lu Tian

Y

Yuzhou Liu

Y

Yazhe Xie

Y

Yinhaibing Ding

DOI：10.1016/j.rse.2026.115612

Original

Original request for help

Save

Abstract

En 中文

• A robust method is proposed for high-resolution C/S-band SAR image matching. • CS-Bench (9 k) provides the first accurately co-registered C/S-band SAR benchmark. • PSSG generates 450 k pseudo-S-band samples to alleviate the scarcity of real paired data. • CSSMatch couples FAM-Conv and MGA to handle scattering differences and speckle noise. • CSSMatch achieves the best matching accuracy across diverse scenes and conditions.

Keywords:

Synthetic aperture radar

High-resolution image matching

GF-3 A/B/C

HJ-2E/F

Deep learning

A robust method for high-resolution C/S-band SAR image matching via pseudo-modality generation and coupled spectrum-spatial learning

Abstract

Journal

Organization

Cited Papers

Citing Papers