POLySight

Towards Practical Bi-Static Polarimetric Imaging Using Commodity mmWave Radars for Material Sensing

ACM/IEEE SenSys 2026

Project Page | Dataset (Hugging Face)

Polarimetric bistatic SAR imaging system for contactless material characterization at 77 GHz.

A single TX board (TI AWR2243) illuminates the scene while two spatially separated RX boards capture orthogonal polarizations (H, V). A 2-D motorized raster stage synthesizes a large aperture. The polarimetric ratio H/V, combined with Fresnel reflection inversion, yields complex permittivity (ε′ − jε″) of materials without physical contact.

Setup

uv venv
uv pip install -r requirements.txt

GPU acceleration (SAR generation) requires CUDA 12.x and CuPy.

Pipeline

Collection → Decode/Sync → SAR Generation → Labeling → Permittivity Extraction
(collect.py)   (decode.py)   (sar.py + GPU)   (notebook)   (Fresnel inversion)

Detailed tutorials/implementations are inside the notebooks folder:

1. Synchronization
2. SAR generation
3. Liquid materials
4. Solid and powder materials
5. Case study on sugar water concentration

Method

1. Bistatic geometry: TX and RX separated ~1.5 m, incidence angle 55°–60°
2. Polarimetric ratio: div = SAR_H / SAR_V (complex-valued)
3. Amplitude → Fresnel ratio: Normalize against metal reference, correct for antenna axial ratio, compute p = tan(θ_pol)
4. Phase → loss tangent: Extract referenced phase from H/V ratio
5. Complex Fresnel ratio: p_complex = |p| · exp(−jφ)
6. Inversion: ε = (1 + 4p/(1−p)² · sin²θ_i) · tan²θ_i
7. Result: ε′ = Re(ε), ε″ = |Im(ε)|

Dataset

The dataset contains 70 synchronized radar collections across 36 materials including plastics, ceramics, glass, wood, rubber, metals, liquids, and powders, measured at 55° and 60° incidence angles.

Folder	Size	Contents
raw/	~72 GB	70 synchronized radar cubes (~1 GB each)
labels/	~66 MB	84 SAR image collections with labels
materials/	~1 MB	72 cropped material sample files (.pkl)
unsynced/	~35 GB	2 example raw unsynchronized ADC captures

See data/README.md for more details and selective downloading.

Hardware

See hardware/README.md for more details

• Radar: 3 sets of TI AWR2243 FMCW (76–81 GHz) + DCA1000 EVM (ADC capture). Currently requires 3 PCs to config and re-trigger each of the radar with TI mmWaveStudio. todo: command line tool to replace the mmWaveStudio. A similar setup can be found in our MulDar repo: https://github.com/xsun2445/MulDar
• Synchronization: Raspberry Pi 4B
• Motion stage: 2-axis linear stage, 3 stepper motors + drivers, Arduino-controlled stepper motors

For more details please check out our paper: https://xsun2445.github.io/assets/pdf/Sensys_2026_PolySight.pdf

Citation

@inproceedings{sun2026towards,
  title={Towards Practical Bi-Static Polarimetric Imaging Using Commodity mmWave Radars for Material Sensing},
  author={Sun, Xinghua and Gadre, Akshay},
  booktitle={Proceedings of the 2026 ACM/IEEE International Conference on Embedded Artificial Intelligence and Sensing Systems},
  pages={172--185},
  year={2026}
}