High-Resolution ISAR Imaging with Modified Joint Range Spatial-Variant Autofocus and Azimuth Scaling

Jiaqi Wei, Shuai Shao, Hui Ma, Penghui Wang, Lei Zhang, Hongwei Liu2020 · DOI

TL;DR

This paper introduces a novel algorithm, MJAAS, for high-resolution Inverse Synthetic Aperture Radar (ISAR) imaging. MJAAS jointly addresses the challenges of range spatial-variant autofocus and azimuth scaling, which are often degraded by motion compensation steps. By accurately estimating the target's equivalent rotational center and velocity, MJAAS significantly improves ISAR image quality and accuracy.

What Problem Does This Paper Address?

Inverse Synthetic Aperture Radar (ISAR) is crucial for identifying and recognizing targets by creating detailed 2D images. However, achieving high-resolution and accurate ISAR images is challenging due to target motion. Motion compensation techniques are applied to correct for these movements, but they can inadvertently shift the target's "equivalent rotational center" (ERC). This shift disrupts the relationship between the azimuth chirp rates (ACR) and the scattering points' positions, making subsequent azimuth scaling (determining the target's true size) inaccurate. Furthermore, existing methods often overlook range spatial-variant phase errors, leading to blurred images. This paper aims to solve these issues by developing an ISAR imaging algorithm that can simultaneously perform accurate azimuth scaling and range spatial-variant autofocus, even after motion compensation.

Methods

The core of this paper is the proposed Modified Joint Range Spatial-Variant Autofocus and Azimuth Scaling (MJAAS) algorithm. MJAAS builds upon a new signal model called the Joint Equivalent Rotational Center Position and Effective Rotational Velocity (JERCP-ERV) model. This model explicitly accounts for the ERC shift caused by motion compensation and jointly estimates the Equivalent Rotational Center Position (ERCP) and Effective Rotational Velocity (ERV).

The estimation of ERCP and ERV is achieved by minimizing the image entropy (IE), a common metric for image quality. The Davidon-Fletcher-Powell (DFP) algorithm, a quasi-Newton method, is employed to solve this minimum entropy optimization problem. The paper also integrates MJAAS with standard motion compensation techniques like Improved Maximum Cross-Correlation (IMCM) for range alignment and Phase Gradient Autofocus (PGA) for phase adjustment, forming a complete ISAR imaging framework.

Datasets

The paper utilizes both simulated and real-world datasets for evaluation.

  • Simulated Data: A 3D scattering point model of a Yak-42 airplane was used to generate simulated ISAR data. This data was subjected to both coherent (CM) and non-coherent (NCM) motion errors and varying Signal-to-Noise Ratios (SNRs).
  • Real Data: Real ISAR data from a Yak-42 airplane and a ship were used to validate the algorithm's performance in practical scenarios.

Key Insights and Ideas

  • ERC Shift is Critical: The paper highlights that standard motion compensation techniques, particularly PGA, can shift the target's Equivalent Rotational Center (ERC). This shift is a primary cause of inaccurate azimuth scaling and degraded autofocus.
  • Joint Estimation is Key: MJAAS's novelty lies in its ability to jointly estimate both the ERCP and ERV. This simultaneous estimation allows for the correction of errors introduced by the ERC shift, leading to more accurate azimuth scaling and autofocus.
  • Image Entropy as an Objective Function: Using image entropy as the objective function for optimization is a well-established technique for image focusing. MJAAS leverages this by minimizing entropy to find the optimal ERCP and ERV.
  • Robustness to Motion Error Modes: The proposed MJAAS algorithm, when integrated into the IMCM-PGA-MJAAS framework, demonstrates robustness to both coherent and non-coherent motion errors, a significant advantage over some existing methods.

Key Results

  • Improved Image Quality: Experimental results on both simulated and real data show that the IMCM-PGA-MJAAS framework produces well-focused and accurately scaled ISAR images, outperforming traditional methods like IMCM-PGA and IMCM-PGA-TAS, especially at lower SNRs.
  • Accurate Azimuth Scaling: The proposed method significantly reduces errors in azimuth scaling compared to methods that do not account for ERC deviation. For instance, real data experiments showed that the estimated target dimensions were much closer to the actual dimensions.
  • Higher Accuracy in ERV Estimation: MJAAS consistently provided more accurate estimates of the Effective Rotational Velocity (ERV) compared to the traditional Azimuth Chirp Rate Estimation Method (ACREM) used in TAS, as evidenced by lower Estimated Correct Rate (ECR) values.
  • Lower Image Entropy: Across various SNRs and motion error modes, MJAAS consistently achieved lower image entropy, indicating superior focusing performance.
  • Efficiency: While not explicitly quantified in terms of speedup, the paper mentions that the framework has "low computational complexity" and "high efficiency."

Limitations and Future Work

  • Maneuvering Targets: The authors acknowledge that more efforts will be devoted to improving the algorithm's performance for "strong maneuvering targets with complex motions."
  • Artificial Intelligence Integration: The paper suggests exploring the integration of artificial intelligence (AI) in future work to further enhance ISAR imaging precision, robustness, and computational efficiency.
  • Computational Complexity: While stated as low, the paper does provide a complexity analysis of O(U · (2 · (V + 1) · N · (M log2M))), where U and V are iterative numbers. Further optimization might be explored.

Summary updated

BibTeX
@misc{wei2020high,
  author = {Jiaqi Wei and Shuai Shao and Hui Ma and Penghui Wang and Lei Zhang and Hongwei Liu},
  title = {{High-Resolution ISAR Imaging with Modified Joint Range Spatial-Variant Autofocus and Azimuth Scaling}},
  year = {2020},
  doi = {10.3390/s20185047},
}

Loading related content...

Figure