Detection of Unresolved Targets for Wideband Monopulse Radar

Tianyi Tsai, Zhiqiang Liao, Zhiquan Ding, Yuan Zhao, Bin TangSensors 2019 · DOI

TL;DR

This paper introduces a new algorithm, GBD (Gaussian Mixture Model-based Bayesian Detector), for detecting unresolved targets in wideband monopulse radars. GBD uses a Gaussian Mixture Model to adapt to various radar resolutions and incorporates prior tracking information to achieve significantly higher detection probabilities and much lower computational costs compared to existing methods like JMBP GLRT and Blair GLRT.

What Problem Does This Paper Address?

Detecting multiple targets that are too close to be distinguished by the radar (unresolved targets) is a critical challenge in radar systems, especially for target tracking. In wideband radars, existing algorithms often struggle with adaptability to different bandwidths and can be computationally expensive. Traditional methods often rely on assumptions about echo distributions (like Rayleigh) that break down with increasing radar resolution. This paper aims to develop a fast, adaptive, and high-performance algorithm for detecting these unresolved targets in wideband monopulse radars, overcoming the limitations of previous approaches.

Methods

The core of the proposed GBD algorithm lies in two key innovations:

  1. Gaussian Mixture Model (GMM) for Echo Modeling: Instead of assuming specific echo distributions, GBD uses a GMM to model the probability density function (PDF) of the monopulse ratio. This allows the algorithm to learn and adapt to the actual echo characteristics, making it more robust to different radar resolutions and target structures.
  2. Bayesian Detection with Prior Information: GBD leverages the detection results from previous radar processing periods. By treating unresolved targets as a special "u-target" and employing a Kalman filter for tracking, it derives a priori probabilities for current detections. This prior information is then integrated into a Bayesian framework to improve detection accuracy and reduce false alarms.

The algorithm operates in two modes: one using only GMM when target tracking is not yet established, and another using both GMM and prior information when tracking is active.

Datasets

The paper uses two types of data for evaluation:

  • Simulated Data: The authors generated simulated radar echoes with Rayleigh and bimodal distributions to compare GBD against JMBP GLRT under various Signal-to-Noise Ratios (SNRs) and angular separations.
  • Measured Field Test Data: Real-world data collected from a wideband monopulse radar experiment involving a ship and a moving corner reflector was used to evaluate GBD against Blair GLRT. This data was used to train the GMM models for both normal ship echoes and unresolved target scenarios.

Key Insights and Ideas

  • Adaptability is Key: The GMM approach is a significant departure from fixed distribution assumptions, allowing the algorithm to generalize across different radar resolutions and target complexities. This is crucial as radar technology advances towards higher resolutions.
  • Leveraging Past Information: Incorporating tracking information from previous time steps as a priori probabilities dramatically boosts detection performance. This highlights the value of temporal correlation in radar signal processing.
  • Computational Efficiency: The GBD algorithm achieves remarkable speedups (orders of magnitude) over methods like JMBP GLRT, which rely on computationally intensive Maximum Likelihood (ML) estimations. This is critical for real-time radar applications.
  • Unresolved Targets as "u-targets": Conceptualizing unresolved targets as a single, trackable "u-target" simplifies the problem and enables the effective use of tracking filters.

Key Results

  • Simulation (Rayleigh Distribution): GBD achieved up to 0.5456 higher detection probability than JMBP GLRT with computation times that were less than 0.0002 times that of JMBP GLRT.
  • Simulation (Bimodal Distribution): GBD showed up to 0.7933 higher detection probability than JMBP GLRT, with computation times less than 0.0001 times that of JMBP GLRT.
  • Simulation (Multiple Targets): As the number of unresolved targets increased, GBD's detection probability increased while its false alarm rate decreased.
  • Experiment (Measured Data): GBD demonstrated a higher detection probability and lower false alarm rate compared to Blair GLRT. It successfully reduced angular glint caused by unresolved targets, leading to improved angle estimation accuracy. The algorithm completed detections on a range profile within 0.22 ms.

Limitations and Future Work

  • Computational Load: While significantly faster than JMBP GLRT, GBD's computational load is still higher than Blair GLRT due to its tracking and GMM likelihood calculations. The authors suggest that a C language implementation could further reduce this time.
  • Model Mismatch: The paper notes that GBD avoids the risk of model mismatch that Blair GLRT might suffer in wideband radars.
  • Future Work: The authors propose to further explore optimizing the signal model and tracking algorithm to handle more complex real-world scenarios. They also aim to adapt the algorithm for more complicated scenarios beyond the scope of this paper.

Summary updated

Abstract

Detecting unresolved targets is very important for radars in their target tracking phase. For wideband radars, the unresolved target detection algorithm should be fast and adaptive to different bandwidths. To meet the requirements, a detection algorithm for wideband monopulse radars is proposed, which can detect unresolved targets for each range profile sampling points. The algorithm introduces the Gaussian mixture model and uses a priori information to achieve high performance while keeping a low computational load, adaptive to different bandwidths. A comparison between the proposed algorithm and the latest unresolved target detection algorithm Joint Multiple Bin Processing Generalized Likelihood Ratio Test (JMBP GLRT) is carried out by simulation. On Rayleigh distributed echoes, the detection probability of the proposed algorithm is at most 0.5456 higher than the JMBP GLRT for different signal-to-noise ratios (SNRs), while the computation time of the proposed algorithm is no more than two 10,000ths of the JMBP GLRT computation time. On bimodal distributed echoes, the detection probability of the proposed algorithm is at most 0.7933 higher than the JMBP GLRT for different angular separations of two unresolved targets, while the computation time of the proposed algorithm is no more than one 10,000th of the JMBP GLRT computation time. To evaluate the performance of the proposed algorithm in a real wideband radar, an experiment on field test measured data was carried out, in which the proposed algorithm was compared with Blair GLRT. The results show that the proposed algorithm produces a higher detection probability and lower false alarm rate, and completes detections on a range profile within 0.22 ms.
BibTeX
@inproceedings{tsai2019detection,
  author = {Tianyi Tsai and Zhiqiang Liao and Zhiquan Ding and Yuan Zhao and Bin Tang},
  title = {{Detection of Unresolved Targets for Wideband Monopulse Radar}},
  booktitle = {Sensors},
  year = {2019},
  doi = {10.3390/s19051084},
}

Loading related content...

Figure