Abstract:Recent advancements in synthetic aperture radar (SAR) ship detection using deep learning have significantly improved accuracy and speed. However, detecting small targets against complex backgrounds remains a challenge. This letter introduces RSNet, a lightweight framework designed to enhance ship detection in SAR imagery. To ensure accuracy with fewer parameters, RSNet uses Waveletpool-ContextGuided (WCG) as its backbone, guiding global context understanding through multi-scale wavelet features for effective detection in complex scenes. Additionally, Waveletpool-StarFusion (WSF) is introduced as the neck, employing a residual wavelet element-wise multiplication structure to achieve higher-dimensional nonlinear features without increasing network width. The LS module is introduced as detect components to achieve efficient detection through lightweight shared convolutional structure and multi-format compatibility. Experiments on the SAR Ship Detection Dataset (SSDD) and High-Resolution SAR Image Dataset (HRSID) demonstrate that RSNet achieves a strong balance between lightweight design and detection performance, surpassing many state-of-the-art detectors, reaching 72.5% and 67.6% in \textbf{\(\mathbf{mAP_{.50:.95}}\) }respectively with 1.49M parameters. Our code will be released soon.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in complex backgrounds, especially in high - noise environments, how to improve the detection accuracy and speed of small targets (such as ships) in Synthetic Aperture Radar (SAR) images. Traditional methods face challenges when dealing with these complex scenarios, especially when achieving efficient detection while keeping the model lightweight. ### Specific problems: 1. **Small - target detection in complex backgrounds**: There is a large amount of interference information (such as land structures, clutter, and coastlines) in SAR images, which makes it difficult to accurately detect small ships. 2. **Balance between model lightweight and performance**: In order to adapt to the computing and energy limitations of aerial platforms, it is necessary to design a lightweight model framework that maintains high detection performance while reducing the number of parameters. 3. **Multi - scale feature fusion**: Targets in SAR images may appear at different scales, so an effective multi - scale feature fusion mechanism is required to improve the detection effect. ### Solutions: To solve the above problems, the paper proposes RSNet (Restar the Sky Network), a lightweight framework specifically used for multi - scale target detection in SAR images. The main improvements of RSNet include: - **Waveletpool - ContextGuided (WCG) backbone network**: By introducing the wavelet pooling mechanism, it effectively extracts multi - scale spatial features, and at the same time, combines the context - guided convolution module to enhance the detection ability of small targets in complex backgrounds. \[ Y_{\text{pool}}=\text{Conv2D}(Y_{C2f})\in\mathbb{R}^{B\times C_{\text{out}}\times H\times W} \] \[ F = \{F_{LL}, F_{LH}, F_{HL}, F_{HH}\} \] \[ Y_{\text{pool}}=\sum_{j = 1}^4F_j*Y_{C2f} \] - **Waveletpool - StarFusion (WSF) neck structure**: Through wavelet up - sampling and residual element - multiplication structure, it restores the spatial resolution and fuses features, thereby achieving a higher - dimensional non - linear feature representation without increasing the network width. \[ y=x_{\text{in}}+\text{dropout}(x_{\text{new}}) \] - **Lightweight - Shared (LS) detection head**: Use the shared convolution structure and multi - format compatibility to reduce the number of parameters and improve the detection efficiency. Through these improvements, the experimental results of RSNet on the SSDD and HRSID datasets show that while maintaining a lightweight design, it achieves performance superior to many existing detectors, reaching mAP.50:.95 metrics of 72.5% and 67.6% respectively, and the number of parameters is only 1.49M.

RSNet: A Light Framework for The Detection of Multi-scale Remote Sensing Targets

A Lightweight SAR Ship Detector Using End-to-End Image Preprocessing Network and Channel Feature Guided Spatial Pyramid Pooling

SRT-Net: Scattering Region Topology Network for Oriented Ship Detection in Large-Scale SAR Images

DSF-Net: A Dual Feature Shuffle Guided Multi-Field Fusion Network for SAR Small Ship Target Detection

Context-aware SAR image ship detection and recognition network

R-LRBPNet: A Lightweight SAR Image Oriented Ship Detection and Classification Method

Light-SAR-ShipNet (LSSNet): A Lightweight SAR Image Ship Detection System with Enhanced Deep Supervision

LPDNet: A Lightweight Network for SAR Ship Detection Based on Multi-Level Laplacian Denoising

R-CenterNet+: Anchor-Free Detector for Ship Detection in SAR Images

MSSD-Net: Multi-Scale SAR Ship Detection Network

LRMSNet: A New Lightweight Detection Algorithm for Multi-Scale SAR Objects

A Lightweight Faster R-CNN for Ship Detection in SAR Images

LRTransDet: A Real-Time SAR Ship-Detection Network with Lightweight ViT and Multi-Scale Feature Fusion

LMSD-Net: A Lightweight and High-Performance Ship Detection Network for Optical Remote Sensing Images

LSR-Det: A Lightweight Detector for Ship Detection in SAR Images Based on Oriented Bounding Box

Stepwise Attention-Guided Multiscale Fusion Network for Lightweight and High-Accurate SAR Ship Detection

Detection and Recognition of SAR Small Ship Objects Using Deep Neural Network

DEPDet: A Cross-Spatial Multiscale Lightweight Network for Ship Detection of SAR Images in Complex Scenes

Contextual Region-Based Convolutional Neural Network with Multilayer Fusion for SAR Ship Detection

SEFRNet: An SAR Ship Target Detection Network With Effective Feature Representation

CSEF-Net: Cross-Scale SAR Ship Detection Network Based on Efficient Receptive Field and Enhanced Hierarchical Fusion