Near Field 3-D Millimeter-Wave SAR Image Enhancement and Detection with Application of Antenna Pattern Compensation

Shaoqiu Song, Jie Lü, Shiqi Xing, Sinong Quan, Junpeng Wang, Yongzhen Li, Jing LianSensors 2022 · DOI

TL;DR

This paper introduces an "Enhanced Back-Projection Algorithm" (EBPA) for near-field 3D millimeter-wave (mmWave) SAR imaging. It tackles the challenge of image degradation in non-ideal environments and with sparse arrays by incorporating antenna pattern compensation, leading to higher fidelity reconstructions and improved target detection.

What Problem Does This Paper Address?

Millimeter-wave (mmWave) SAR imaging is becoming increasingly important for applications like automotive sensing, UAVs, and threat detection. However, current near-field SAR imaging techniques often struggle in real-world, non-ideal conditions. These challenges include the effects of complex electromagnetic environments and the use of sparse antenna arrays, which can lead to ghosting, clutter, and reduced image quality. Existing algorithms like the standard Back-Projection Algorithm (BPA) and Range Migration Algorithm (RMA) are often limited in their ability to handle these non-ideal scenarios, particularly when dealing with the spherical wave propagation in near-field conditions. This paper aims to bridge this gap by proposing a method that enhances image reconstruction quality and robustness in these challenging environments.

Methods

The core of this paper is the development of an Enhanced Back-Projection Algorithm (EBPA). This method builds upon the traditional BPA by incorporating beam-weighting amplitude compensation. Instead of simply using the amplitude factor in the echo model, the EBPA converts it into a beam-weighting factor. This approach is designed to:

  • Compensate for antenna pattern variations: By considering the antenna's directional response, the algorithm can better account for how signals are transmitted and received.
  • Mitigate ghosting and clutter: This is particularly important when dealing with sparse arrays where antenna spacing might exceed half a wavelength, leading to grating lobes.
  • Improve image focusing: The EBPA aims for a higher fidelity reconstruction by coherently accumulating signals with this enhanced compensation.

The paper also discusses the underlying signal model for FMCW radar and briefly touches upon the Range Migration Algorithm (RMA) with amplitude loss compensation as a point of comparison. For target detection, the YOLO (You Only Look Once) network is employed, leveraging the improved image quality from the EBPA.

Datasets

The paper utilizes the 3DRIED dataset for constructing image datasets for target detection. This dataset appears to be a benchmark specifically designed for near-field SAR imaging scenarios, allowing for the evaluation of imaging algorithms and subsequent detection performance.

Key Insights and Ideas

  • Antenna Pattern Compensation is Crucial for Near-Field SAR: The paper strongly suggests that simply applying standard imaging algorithms without accounting for the antenna's spatial characteristics leads to significant degradation in near-field scenarios, especially with sparse arrays.
  • Beam-Weighting as a Solution to Ghosting: The EBPA's novel approach of converting amplitude factors to beam-weighting effectively suppresses ghost images and clutter, which are persistent problems in traditional BPA and RMA.
  • Real-World Performance Matters: The research highlights that laboratory-confined studies are insufficient; the true test of an algorithm lies in its performance under non-ideal, complex environments.
  • Improved Image Quality Directly Benefits Detection: The study demonstrates a clear link between the enhanced image reconstruction quality provided by EBPA and improved target detection accuracy using YOLO.

Key Results

  • Superior Image Reconstruction: The EBPA consistently produced higher fidelity images compared to classic BPA and RMA with amplitude loss compensation, particularly evident in suppressing side lobes and grating lobes.
  • Quantitative Improvements:
    • Image Contrast: The EBPA showed a higher image contrast (e.g., 134.5673 for scissor vs. 89.5611 for BPA and 112.6986 for RMA).
  • Image Entropy: The EBPA achieved lower image entropy (e.g., 3.8054 for scissor vs. 4.2688 for BPA and 3.8837 for RMA), indicating better focusing.
  • Enhanced Target Detection: The kappa coefficients for classification accuracy were improved with the EBPA, demonstrating its effectiveness in providing cleaner inputs for detection algorithms. For instance, the kappa coefficient for "knife" was 0.92 with EBPA, compared to 0.87 for BPA and 0.85 for RMA.
  • Ghost Suppression: The paper visually demonstrates significant reduction in ghosting artifacts, especially in scenarios with sparse antenna arrays.

Limitations and Future Work

  • Computational Complexity: While the EBPA offers improved performance, the paper acknowledges that its computational complexity is still enormous. They suggest using parallel processing techniques (like MATLAB's parfor or GPU acceleration) to speed up computation.
  • Generalizability: While the paper demonstrates effectiveness in complex environments, further validation across a wider range of diverse scenarios and targets would be beneficial.
  • Further Optimization: The authors aim to contribute a complete and efficient imaging and detection process, suggesting that ongoing work might focus on further optimizing both the imaging and detection stages for even greater efficiency and accuracy.

Summary updated

Abstract

In this paper, a novel near-field high-resolution image focusing technique is proposed. With the emergence of Millimeter-wave (mmWave) devices, near-field synthetic aperture radar (SAR) imaging is widely used in automotive-mounted SAR imaging, UAV imaging, concealed threat detection, etc. Current research is mainly confined to the laboratory environment, thus ignoring the adverse effects of the non-ideal experimental environment on imaging and subsequent detection in real scenarios. To address this problem, we propose an optimized Back-Projection Algorithm (BPA) that considers the loss path of signal propagation among space by converting the amplitude factor in the echo model into a beam-weighting. The proposed algorithm is an image focusing algorithm for arbitrary and irregular arrays, and effectively mitigates sparse array imaging ghosts. We apply the 3DRIED dataset to construct image datasets for target detection, comparing the kappa coefficients of the proposed scheme with those obtained from classic BPA and Range Migration Algorithm (RMA) with amplitude loss compensation. The results show that the proposed algorithm attains a high-fidelity image reconstruction focus.
BibTeX
@inproceedings{song2022near,
  author = {Shaoqiu Song and Jie Lü and Shiqi Xing and Sinong Quan and Junpeng Wang and Yongzhen Li and Jing Lian},
  title = {{Near Field 3-D Millimeter-Wave SAR Image Enhancement and Detection with Application of Antenna Pattern Compensation}},
  booktitle = {Sensors},
  year = {2022},
  doi = {10.3390/s22124509},
}
Figure