Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IFNet: Deep Imaging and Focusing for Handheld SAR with Millimeter-wave Signals

Yadong Li, Dongheng Zhang, Ruixu Geng, Jincheng Wu, Yang Hu, Qibin Sun, Yan Chen

TL;DR

This work tackles the problem of imaging with handheld mmWave SAR where motion induces phase errors that degrade image quality. It proposes IFNet, a deep unfolding network that unifies a physics-based imaging model with a learnable phase error prior, cast as a multi-stage optimization that alternates imaging and focusing. By preserving complex-valued inputs, using a sparse prior for the image, and learning a deep prior for phase compensation, IFNet achieves large gains in PSNR and SSIM on real and simulated datasets, outperforming prior cGAN-based baselines. The results suggest strong potential for practical handheld mmWave imaging applications, and the authors provide dataset and code to facilitate further research in this area.

Abstract

Recent advancements have showcased the potential of handheld millimeter-wave (mmWave) imaging, which applies synthetic aperture radar (SAR) principles in portable settings. However, existing studies addressing handheld motion errors either rely on costly tracking devices or employ simplified imaging models, leading to impractical deployment or limited performance. In this paper, we present IFNet, a novel deep unfolding network that combines the strengths of signal processing models and deep neural networks to achieve robust imaging and focusing for handheld mmWave systems. We first formulate the handheld imaging model by integrating multiple priors about mmWave images and handheld phase errors. Furthermore, we transform the optimization processes into an iterative network structure for improved and efficient imaging performance. Extensive experiments demonstrate that IFNet effectively compensates for handheld phase errors and recovers high-fidelity images from severely distorted signals. In comparison with existing methods, IFNet can achieve at least 11.89 dB improvement in average peak signal-to-noise ratio (PSNR) and 64.91% improvement in average structural similarity index measure (SSIM) on a real-world dataset.

IFNet: Deep Imaging and Focusing for Handheld SAR with Millimeter-wave Signals

TL;DR

This work tackles the problem of imaging with handheld mmWave SAR where motion induces phase errors that degrade image quality. It proposes IFNet, a deep unfolding network that unifies a physics-based imaging model with a learnable phase error prior, cast as a multi-stage optimization that alternates imaging and focusing. By preserving complex-valued inputs, using a sparse prior for the image, and learning a deep prior for phase compensation, IFNet achieves large gains in PSNR and SSIM on real and simulated datasets, outperforming prior cGAN-based baselines. The results suggest strong potential for practical handheld mmWave imaging applications, and the authors provide dataset and code to facilitate further research in this area.

Abstract

Recent advancements have showcased the potential of handheld millimeter-wave (mmWave) imaging, which applies synthetic aperture radar (SAR) principles in portable settings. However, existing studies addressing handheld motion errors either rely on costly tracking devices or employ simplified imaging models, leading to impractical deployment or limited performance. In this paper, we present IFNet, a novel deep unfolding network that combines the strengths of signal processing models and deep neural networks to achieve robust imaging and focusing for handheld mmWave systems. We first formulate the handheld imaging model by integrating multiple priors about mmWave images and handheld phase errors. Furthermore, we transform the optimization processes into an iterative network structure for improved and efficient imaging performance. Extensive experiments demonstrate that IFNet effectively compensates for handheld phase errors and recovers high-fidelity images from severely distorted signals. In comparison with existing methods, IFNet can achieve at least 11.89 dB improvement in average peak signal-to-noise ratio (PSNR) and 64.91% improvement in average structural similarity index measure (SSIM) on a real-world dataset.
Paper Structure (24 sections, 14 equations, 12 figures, 1 table)

This paper contains 24 sections, 14 equations, 12 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related work
  3. Handheld mmWave Imaging
  4. Autofocus
  5. Deep Unfolding Network
  6. Preliminary and challenges
  7. MIMO SAR Imaging Fundamentals
  8. Analysis of Handheld Motion Errors
  9. Method
  10. Imaging and Focusing with Sparse and Deep Prior
  11. Imaging and Focusing with Deep Unfolding Network
  12. Experiments
  13. Implementation Details
  14. Evaluation Datasets
  15. Real measurements with 26 letters
  16. ...and 9 more sections

Figures (12)

  • Figure 1: Handheld SAR imaging has portable applications and IFNet is designed to reconstruct images distorted by handheld motion errors.
  • Figure 2: Imaging model of MIMO SAR.
  • Figure 3: Impact of handheld motion errors. Even sub-mm scale motion errors can lead to significant image distortion.
  • Figure 4: The architecture of IFNet. IFNet comprises multiple stages, with each stage consisting of an imaging module and a focusing module. The design of the imaging and focusing modules are based on the optimization-based signal model.
  • Figure 5: Collecting ground truths with mechanical scanning and handheld scanning patterns with tracking camera.
  • ...and 7 more figures