ADAPTIVE CLUTTER SUPPRESSION FOR AIRBORNE RANDOM PULSE REPETITION INTERVAL RADAR BASED ON COMPRESSED SENSING

Zhen Liu, Xizhang Wei, Xiang LiElectromagnetic 2012 · DOI

Abstract

We present an adaptive clutter suppression method for airborne random pulse repetition interval radar by using prior knowledge of clutter boundary in Doppler spectrum. In this method, by exploiting the intrinsic sparsity, compressed sensing based on iterative grid optimization (CS-IGO) is applied to directly recover the clutter spectrum with only the test range cell instead of nonhomogeneous training data from adjacent range cells. Since the sensing matrix and clutter spectrum obtained by CS-IGO are well adapted to the data, the prewhitening filter can be effectively obtained to cancel the mainlobe clutter. Further, the clutter residue can be suppressed by the iterative reweighted $l_1$ minimization to enhance the target response. Simulation results show that the approach is capable of effective suppression of clutter and precise recovery of targets' unambiguous spectrum, offering a high performance of output signal to clutter and noise ratio.

Introduction and Problem Statement

Airborne radar systems face a significant challenge in detecting moving targets due to heavy ground clutter. When a radar platform looks down from an aircraft, it receives massive reflections from the terrain—mountains, buildings, and the ground—which can easily drown out the relatively weak signals from moving targets.

Modern radar systems often employ Random Pulse Repetition Intervals (PRI) to enhance electronic counter-countermeasures (ECCM) and resolve velocity ambiguities. However, this randomization introduces a new problem: it creates high Doppler sidelobes, essentially "ghost signals" that mask real targets. Traditional Space-Time Adaptive Processing (STAP) techniques attempt to solve this by using "training data" from adjacent range cells to estimate the clutter environment.

Geometry of the airborne radar system illustrating the relationship between the test range cell and adjacent range cells.

As shown in the geometry above, traditional methods rely on the assumption that adjacent range cells share the same statistical properties as the test range cell. In reality, if the terrain changes quickly (nonhomogeneous clutter), these adjacent areas do not accurately represent the test cell, leading to filter failure. This research proposes a "Direct Data Domain" approach, suppressing clutter using only the data from the specific test range cell being analyzed.

Proposed Methodology

The authors introduce a three-stage pipeline based on the principles of Compressed Sensing (CS). This approach exploits the inherent sparsity of the radar scene, where only a few targets and a localized clutter band exist within a much larger Doppler space.

1. CS-IGO (Iterative Grid Optimization)

Standard CS methods often struggle with "off-grid" targets—signals whose velocities do not align perfectly with a predefined mathematical grid. This leads to spectral leakage and estimation errors. The proposed CS-IGO algorithm iteratively adjusts the grid to match the actual data, allowing for a high-resolution estimation of the clutter's power spectral density without requiring training data from neighboring cells.

2. Adaptive Prewhitening Filter

Once the high-resolution clutter spectrum is estimated, the system constructs an adaptive prewhitening filter. This filter is designed to "cancel out" the mainlobe clutter—the strongest part of the interference—by projecting the received signal into a subspace that is theoretically clutter-free.

3. Iterative Reweighted $l_1$ Minimization

Even after prewhitening, some clutter residue may remain. To address this, the authors apply a final CS recovery step using iterative reweighted $l_1$ minimization. This technique uses weights to penalize small, noisy components (the residue) while encouraging the recovery of strong, clear target signals, effectively sharpening the final radar output.

Experimental Setup and Results

The effectiveness of the proposed method was evaluated through numerical simulations using the following parameters:

  • Radar Band: X-band (0.03 m wavelength).
  • Signal Specs: 64 pulses per coherent processing interval (CPI) of 6.4 ms.
  • Clutter Model: 50 independent clutter scatters placed within specific Doppler cells (31st–40th), modeled with Gaussian complex amplitudes.
  • Environment: Simulated in MATLAB using the CVX toolbox for convex optimization.

Key Findings

  • Superior Clutter Suppression: The CS-IGO approach produced significantly fewer false components compared to standard CS methods, which are prone to grid-mismatch errors.
  • Enhanced Signal Quality: The method achieved a much higher Output SCNR (Signal-to-Clutter-and-Noise Ratio) than traditional weighting methods, particularly in scenarios with weak input signals.
  • High Precision: The Normalized Square Error (NSE) for target amplitude was consistently lower, indicating that the system can measure a target's size and velocity with high accuracy.
  • Automation: Unlike previous weighting methods that required manual tuning of scale factors, this approach is largely automated and adapts dynamically to the environment.

Discussion and Conclusion

The primary advantage of this research is the elimination of the need for adjacent training data, making the radar robust in rapidly changing environments, such as transitions from forest to urban terrain. By treating the grid as an optimizable variable, the authors achieve estimation accuracy close to the theoretical Cramer-Rao Lower Bound.

However, these gains come with a computational trade-off. The authors note that the method is more time-consuming than traditional algorithms because it requires solving multiple convex optimization problems. Future work will focus on implementing "fast $l_1$ minimization" algorithms to meet real-time processing requirements and extending the framework from narrowband to wideband imaging radar systems.

Summary updated

Abstract

We present an adaptive clutter suppression method for airborne random pulse repetition interval radar by using prior knowledge of clutter boundary in Doppler spectrum. In this method, by exploiting the intrinsic sparsity, compressed sensing based on iterative grid optimization (CS-IGO) is applied to directly recover the clutter spectrum with only the test range cell instead of nonhomogeneous training data from adjacent range cells. Since the sensing matrix and clutter spectrum obtained by CS-IGO are well adapted to the data, the prewhitening filter can be effectively obtained to cancel the mainlobe clutter. Further, the clutter residue can be suppressed by the iterative reweighted l 1 minimization to enhance the target response. Simulation results show that the approach is capable of effective suppression of clutter and precise recovery of targets' unambiguous spectrum, offering a high performance of output signal to clutter and noise ratio.
BibTeX
@inproceedings{liu2012adaptive,
  author = {Zhen Liu and Xizhang Wei and Xiang Li},
  title = {{ADAPTIVE CLUTTER SUPPRESSION FOR AIRBORNE RANDOM PULSE REPETITION INTERVAL RADAR BASED ON COMPRESSED SENSING}},
  booktitle = {Electromagnetic},
  year = {2012},
  doi = {10.2528/pier12022001},
}
Figure