🚀 Mamba-Based Time–Frequency Interactive Fusion Network for SAR Ship Wake Detection (IEEE TAES 2026)

Lu Wang^1,2, Bailiang Sun^1,✉, Hongxing Wang³, Shaoguang Huang⁴, Tomoaki Ohtsuki⁵, P. Takis Mathiopoulos^1,2,6

¹Harbin Engineering University, ²Ministry of Industry and Information Technology, ³Chongqing University, ⁴China University of Geosciences, ⁵Keio University, ⁶National and Kapodistrian University of Athens

🧠 Abstract

SAR ship wakes exhibit elongated structures and strong directionality, making their detection challenging under sea clutter and speckle noise. TFIFNet combines Mamba-based long-range modeling with time–frequency feature interaction to capture wake morphology and structural continuity.

TFIFNet enhances complementary time–frequency representations through three key components:

• Cross-domain Feature Interaction Module (CFIM): models complementary relationships between time-domain and frequency-domain wake features.
• Complementarity-Guided Dense Fusion (CGDF): integrates cross-domain features using global complementary information.
• Fusion-Guided Feature Enhancement (FGFE): strengthens the discriminative representations of both branches using fused features.

🏗️ Model Overview

⚙️ Installation

Validated environment:

• Python 3.9
• CUDA 12.4
• PyTorch 2.0.1

# 1) Create environment
conda create -n tfifnet python=3.9 -y
conda activate tfifnet

# 2) Install PyTorch
pip install torch==2.0.1 torchvision==0.16.2 \
  --index-url https://download.pytorch.org/whl/cu118

# 3) Clone this repository
git clone https://github.com/luwang-Lucia/TFIFNet.git
cd TFIFNet

# 4) Build modules for calculating OBB overlap and performing Radon transform:
cd $ROOT/utils
sh make.sht

📚 Citation

If you find TFIFNet useful in your research or applications, please consider giving this repository a star ⭐ and citing our paper.

@article{wang2026tfifnet,
  title   = {Mamba-Based Time--Frequency Interactive Fusion Network for SAR Ship Wake Detection},
  author  = {Wang, L. and Sun, B. and Wang, H. and Huang, S. and Ohtsuki, T. and Mathiopoulos, P. T.},
  journal = {IEEE Transactions on Aerospace and Electronic Systems},
  year    = {2026},
  volume  = {62},
   pages  = {16327-16342},
  url     = {https://ieeexplore.ieee.org/document/11664306/}
}