mmUSAR：Millimeter-Wave SAR imaging of Sparse Trajectory via Untrained Complex-valued Neural Network

Abstract: Millimeter-wave Synthetic Aperture Radar (SAR) imaging is widely used in security screening, industrial nondestructive testing, and autonomous navigation due to its high resolution and penetration capabilities. However, high-fidelity reconstruction typically requires dense spatial sampling, resulting in long scanning times and significant data acquisition overhead. Although compressed sensing has been explored for sparse reconstruction, its reliance on random sampling is incompatible with the deterministic scanning trajectories of practical SAR systems and thus fails to reduce real acquisition time. To address this, we propose mmUSAR, an unsupervised SAR imaging framework incorporating two hardware-compliant sparse scanning strategies: Expanded Vertical-interval Sampling (EVS) and Reduced Height Sampling (RHS), which reduce the number of vertical scanning positions. Reconstruction is performed by an untrained complexvalued neural network that embeds physical priors via an observation approximation operator and enforces structured sparsity regularization. The framework requires no external training data and adapts to each scene through test-time optimization. Experiments on simulated and real-world datasets show that mmUSAR achieves high-fidelity image reconstruction while reducing the scanning trajectory, thereby decreasing acquisition time by up to 75%. Compared to conventional and other untrained neural network methods, mmUSAR offers superior image quality and competitive runtime, providing a practical solution for efficient and resource constrained SAR imaging.

✨ Key Features

• Unsupervised Learning: No need for large amounts of labeled training data
• Complex-valued Neural Networks: Native support for complex radar data processing
• Physics-informed Architecture: Integrates radar physical models into neural networks
• Sparse Sampling Support: High-quality reconstruction under sparse trajectory constraints

🚀 Quick Start

Prerequisites

We recommend using conda to create a virtual environment for this project:

# Create conda environment
conda create -n mmusar python=3.8

# Activate environment
conda activate mmusar

# Install dependencies
pip install -r requirements.txt

Data Preparation

Important: Due to the large size of radar datasets, please download the real measurement data manually:

1. Visit 3DRIED Datasets
2. Download the dataset and place it in the datasets/3DRIED/ directory
3. Ensure the data structure follows the expected format as shown in the project layout

Running Experiments

1. Real Data Experiment:

   jupyter notebook main_real.ipynb

2. Simulation Data Experiment:

   jupyter notebook main_sim.ipynb

Note: The notebook approach provides an interactive environment that makes it easier to visualize results and adjust parameters compared to traditional Python scripts.

🔖 Result Display

---

Real Target	mmUSAR Reconstruction Process

---

---

---

---

📖 Citation

If you use this code in your research, please cite our paper:

@ARTICLE{11189855,
  author={Hu, Tingkai and Xu, Junjie and Li, Yipeng and Xiong, Hailing and Luo, Zhen and Li, Chuandong},
  journal={IEEE Transactions on Aerospace and Electronic Systems}, 
  title={Millimeter-Wave SAR imaging of Sparse Trajectory via Untrained Complex-valued Neural Network}, 
  year={2025},
  volume={},
  number={},
  pages={1-15},
  keywords={Radar polarimetry;Neural networks;Image reconstruction;Trajectory;Radar imaging;Radar;Synthetic aperture radar;Imaging;Image resolution;Reflectivity;Synthetic Aperture Radar imaging;Millimeter Wave;Sparse Trajectory;Untrained Neural Network;Unsupervised Reconstruction},
  doi={10.1109/TAES.2025.3617036}
}

🙏 Acknowledgement

[1] 3DRIED Datasets