⚡ electrical engineering

Moving Target SAR Imaging Using Planar Arrays And Multidimensional Chinese Remainder Theorem (MD-CRT)--Part I: A General Framework

This paper presents a general framework for moving target SAR imaging using planar arrays that unifies motion-induced cross-range shift and target height estimation into a vector remainder formulation, resolving ambiguity through a multi-subarray approach based on the multidimensional Chinese Remainder Theorem (MD-CRT) with robust error bounds for practical noisy conditions.

Original authors: Guangpu Guo, Xiang-Gen Xia

Published 2026-06-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Guangpu Guo, Xiang-Gen Xia

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to take a picture of a speeding car driving over a bumpy, three-dimensional mountain road using a special radar camera. This isn't just a normal photo; it's a Synthetic Aperture Radar (SAR) image, which works like a super-powered flashlight that can see through clouds and darkness.

Here is the problem: When the car moves, the radar gets confused. It sees the car in the wrong spot (a "ghost" image) because the motion shifts the signal. Furthermore, because the road is bumpy (3D terrain), the radar also struggles to tell exactly how high the car is. It's like trying to guess both the car's speed and its altitude at the same time, but your ruler is broken.

This paper, Part I of a two-part series, proposes a clever new way to fix this confusion using a specific type of radar antenna called a Planar Array. Think of a planar array not as a single line of antennas, but as a flat, 2D grid (like a chessboard) of sensors.

Here is the breakdown of their solution, explained simply:

1. The "Folded Map" Problem

When the radar looks at the moving car, it calculates a "shift" (how far off the car looks) and a "height." However, the math the radar uses is like a circular clock. If the car moves too fast or is too high, the measurement "wraps around" the clock face.

Analogy: Imagine a clock that only goes from 1 to 12. If the car actually moved 13 hours, the clock still just says "1." The radar sees "1" but doesn't know if it's 1, 13, 25, or 37. This is called ambiguity. The radar has a "folded" map where the true location is hidden.

2. The "Chessboard" Advantage

Previous methods used a linear array (a single line of antennas), which is like trying to solve a 3D puzzle with a 1D ruler. It can only measure one direction well.

The Innovation: This paper uses a planar array (a 2D grid). This is like having a 2D ruler that can measure both the "shift" and the "height" simultaneously.
The Math Trick: They use a mathematical tool called a 2D Discrete Fourier Transform (2D-DFT). Think of this as a special lens that turns the messy radar signals into a clear pattern. However, even with this lens, the "clock wrapping" (ambiguity) still happens.

3. The "Chinese Remainder Theorem" (The Master Key)

To fix the "clock wrapping" problem, the authors use a mathematical concept called the Chinese Remainder Theorem (CRT).

The Analogy: Imagine you have three different clocks, but they all run at different speeds and have different numbers of hours (e.g., one has 12 hours, another has 17, another has 19).
- Clock A says: "It's 3 o'clock." (Could be 3, 15, 27...)
- Clock B says: "It's 5 o'clock." (Could be 5, 22, 39...)
- Clock C says: "It's 7 o'clock."
- By combining these three different "folded" answers, you can mathematically figure out the one true time that fits all of them, even if the real time is way outside the range of a single clock.

In this paper, they use multiple sub-arrays (multiple "clocks" with different grid patterns). Each sub-array gives a slightly different "folded" answer. By combining them using the Multidimensional Chinese Remainder Theorem (MD-CRT), they can reconstruct the true speed and height of the car, even if it's moving very fast or is very high up.

4. Dealing with "Noise" and "Rounding Errors"

In the real world, radar signals aren't perfect. They have static (noise) and the math involves rounding numbers (quantization).

The Challenge: If you round a number slightly wrong, and then try to use the "Master Key" (CRT) to solve the puzzle, you might end up with a completely wrong answer. It's like if one of your clocks was off by a few minutes; the final time calculation could be hours off.
The Solution: The authors developed a Robust version of the math. They created a safety net that accounts for these small errors. They proved mathematically that as long as the errors aren't too big, the system can still find the correct location. They also figured out the best way to arrange the antennas on the grid to make this system as resistant to errors as possible.

Summary of What They Claim

The Setup: They created a framework to image moving targets in 3D space using a flat grid of antennas.
The Method: They combine the data from multiple antenna grids using a 2D mathematical transform and the Multidimensional Chinese Remainder Theorem.
The Result: This method creates a much larger "unambiguous range" (a bigger area where the radar can see the target clearly without getting confused) compared to using just one grid or a single line of antennas.
The Proof: They ran computer simulations showing that with two grids, they could correctly locate a target that a single grid would have misidentified.

Note: This paper (Part I) focuses entirely on the mathematical framework and theory. It sets the stage for Part II, which will likely dive deeper into specific hardware designs and more complex simulations. The paper does not claim to have built a physical device yet, nor does it discuss medical or other non-radar applications. It is purely about solving the math puzzle of radar imaging.

Problem Statement

Synthetic Aperture Radar (SAR) imaging of moving targets over three-dimensional terrain (or in air) presents a dual estimation challenge: accurately determining both the motion-induced cross-range shift and the target height. While conventional Velocity SAR (VSAR) systems using linear antenna arrays can estimate cross-range shifts via Discrete Fourier Transform (DFT), they fail to resolve target height because they only sample spatially along the azimuth direction. Furthermore, when target velocities or heights are large, the phase wrapping inherent in DFT processing leads to ambiguity, restricting the unambiguous range of estimation. Existing methods for resolving ambiguity, such as multi-frequency or non-uniform linear arrays, are primarily designed for one-dimensional (1D) scenarios and do not directly address the coupled 2D nature of cross-range shift and height estimation in 3D environments.

Methodology

The paper proposes a general framework utilizing planar antenna arrays and the Multidimensional Chinese Remainder Theorem (MD-CRT) to jointly estimate the cross-range shift and target height.

Signal Model and 2D-DFT Formulation:
- The authors model the SAR geometry with a platform moving at a constant velocity and a planar antenna array (lying in the $x-z$ plane) with elements arranged on a lattice generated by a matrix $M$ .
- The target is assumed to move with a 3D velocity vector.
- After standard SAR processing (including quadratic phase compensation), the complex image across the planar array exhibits a 2D complex exponential structure.
- The motion-induced cross-range shift and target height are unified into a single 2D parameter vector $\mathbf{g}$ .
- Applying a 2D Discrete Fourier Transform (2D-DFT) across the antenna indices reveals that the parameter vector $\mathbf{g}$ is related to the detected peak via a vector remainder formulation modulo a specific integer matrix $N$ (the 2D-DFT matrix modulus). Specifically, $\mathbf{g} \equiv \mathbf{r} \pmod{N}$ , where $\mathbf{r}$ is the observed vector remainder.
Optimal Matrix Modulus Construction:
- Unlike 1D linear arrays where the modulus is a scalar, the 2D setting involves a matrix modulus $N$ . The structure of $N$ affects how errors in the vector remainder propagate to the estimated parameter vector.
- The paper derives an optimal construction for $N$ under a fixed array geometry ( $M$ ) and a fixed number of antenna elements (fixed $|\det(N)|$ ).
- The optimal choice is $N = \alpha \text{adj}(M^\top)$ , which minimizes the error amplification factor $\|M^{-\top}N^{-1}\|_\infty$ , ensuring that remainder errors are balanced across both dimensions rather than amplified in one direction.
Multi-Subarray Framework for Ambiguity Resolution:
- A single planar array yields only a "folded" estimate (the vector remainder) if the true parameter vector lies outside the unambiguous range defined by the Fundamental Parallelepiped (FPD) of $N$ .
- To resolve this, the authors propose a multi-subarray framework using $J$ subarrays, each with a different lattice generator $M_j$ and corresponding modulus $N_j$ .
- This generates a system of matrix congruences: $\mathbf{g} \equiv \beta_j \mathbf{r}_j \pmod{\beta_j N_j}$ .
- The true parameter vector $\mathbf{g}$ is recovered by solving this system using the MD-CRT, which effectively enlarges the unambiguous range to the FPD of the Least Common Right Multiple (LCRM) of the moduli.
Robust Recovery with Practical Errors:
- The paper addresses practical issues where the frequency vector is non-integer, leading to quantization errors in the 2D-DFT peak detection.
- An approximate 2D-DFT peak model is introduced, replacing non-integer frequency vectors with nearby integer vectors to map the problem back to the standard integer vector remainder framework.
- Robust MD-CRT is incorporated to handle both quantization errors (noiseless case) and additive signal noise.
- The authors establish sufficient conditions for robust recovery and derive explicit reconstruction error bounds for both noiseless and noisy settings. These bounds depend on the scaling factors $\alpha_j$ and the noise levels.

Key Contributions

Unified 2D Formulation: The paper establishes a unified 2D vector remainder formulation for the joint estimation of cross-range shift and target height using planar arrays, extending previous 1D VSAR concepts.
Matrix Modulus Optimization: It provides a theoretical derivation for the optimal 2D-DFT matrix modulus that minimizes error propagation, a structural design freedom unique to the 2D setting.
MD-CRT Framework for SAR: It adapts the MD-CRT to SAR imaging, demonstrating how multiple planar subarrays with different lattice generators can jointly resolve ambiguities and enlarge the unambiguous range beyond what is possible with a single array.
Robustness Analysis: The work introduces an approximate peak model for non-integer frequencies and provides rigorous sufficient conditions and error bounds for robust recovery in the presence of quantization and additive noise.

Results

The paper presents numerical simulations to validate the theoretical framework:

Unambiguous Range Enlargement: A simulation with two co-prime planar subarrays demonstrates that the joint unambiguous range (defined by the LCRM of the subarray moduli) is strictly larger than that of any single subarray.
Localization Accuracy: In a noiseless scenario, the two-subarray framework successfully recovers the target location on the road. In contrast, using a single subarray results in a reconstructed target location shifted away from the true path due to ambiguity.
Error Bounds: The derived error bounds confirm that increasing the scaling factor $\alpha$ (effectively increasing the array size or using zero-padding) reduces the reconstruction error, improving accuracy.

Significance and Claims

The authors claim that this work provides the theoretical foundation for planar array design and performance analysis in moving target SAR imaging.

The framework allows for the joint estimation of 3D motion parameters (shift and height) which is not achievable with conventional 1D linear arrays.
The proposed multi-subarray approach offers a systematic method to enlarge the unambiguous range without requiring complex hardware changes beyond the array geometry.
The paper explicitly states that Part II of this two-part series will investigate specific planar subarray designs and provide more detailed simulations on robustness, indicating that Part I focuses on the general framework, mathematical formulation, and theoretical bounds.

The significance lies in bridging the gap between 1D VSAR techniques and the more complex requirements of 3D terrain imaging, offering a mathematically rigorous approach to ambiguity resolution and error management in SAR systems.