A Lightweight Resonance Network for Small and Slender Ship Detection in SAR Images
IEEE Signal Processing Letters · 2026 · 33 · 3232-3236
DOI: 10.1109/lsp.2026.3717519 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Zhi Yang und 3 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben. Gefördert von: Fujian Special Fund Project for High-Quality Development of Marine and Fishery Industries, Minjiang University.
Abstract (Originalsprache)
Lightweight synthetic aperture radar (SAR) ship detectors often struggle to detect small and slender ships because of geometric mismatch, loss of fine structural details, and redundant feature fusion. We propose a lightweight resonance network (LiRN) to address these limitations. The anisotropic geometric scattering (AGS) module combines orthogonal strip convolutions with parameter-free soft thresholding. The resonance injection fusion (RIF) module preserves locally dominant activations through a parameter-free path and learns their fusion with semantic features. In this context, “resonance” denotes the persistence and reinjection of dominant feature responses. The multi-receptive factorization (MRF) neck reduces feature redundancy through multi-receptive-field depthwise mixing and low-rank pointwise channel mixing. On SSDD and HRSID, LiRN achieves ${\mathbf{AP}_{75}}$ scores of 81.6% and 72.8%, respectively, with only 1.28 M parameters, representing a 58% reduction relative to YOLOv8n. Its measured throughput exceeds 100 FPS on an NVIDIA Tesla P40.
Förderung
- Fujian Special Fund Project for High-Quality Development of Marine and Fishery Industries
- Minjiang University
Zitieren
APA
Yang, Z., He, X., Lin, D., & Liu, C. (2026). A Lightweight Resonance Network for Small and Slender Ship Detection in SAR Images. IEEE Signal Processing Letters, 33, 3232-3236. https://doi.org/10.1109/lsp.2026.3717519
Vancouver
Yang Z, He X, Lin D, Liu C. A Lightweight Resonance Network for Small and Slender Ship Detection in SAR Images. IEEE Signal Processing Letters. 2026;33:3232-3236. doi:10.1109/lsp.2026.3717519
BibTeX
@article{Yang2026,
author = {Yang, Zhi and He, Xuezhi and Lin, Dexing and Liu, Chunhai},
title = {A Lightweight Resonance Network for Small and Slender Ship Detection in SAR Images},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {3232--3236},
doi = {10.1109/lsp.2026.3717519}
}
EUVIMED