Target detection in synthetic aperture radar imagery: a state-of-the-art survey (1804.04719v1)

Abstract: Target detection is the front-end stage in any automatic target recognition system for synthetic aperture radar (SAR) imagery (SAR-ATR). The efficacy of the detector directly impacts the succeeding stages in the SAR-ATR processing chain. There are numerous methods reported in the literature for implementing the detector. We offer an umbrella under which the various research activities in the field are broadly probed and taxonomized. First, a taxonomy for the various detection methods is proposed. Second, the underlying assumptions for different implementation strategies are overviewed. Third, a tabular comparison between careful selections of representative examples is introduced. Finally, a novel discussion is presented, wherein the issues covered include suitability of SAR data models, understanding the multiplicative SAR data models, and two unique perspectives on constant false alarm rate (CFAR) detection: signal processing and pattern recognition. From a signal processing perspective, CFAR is shown to be a finite impulse response band-pass filter. From a statistical pattern recognition perspective, CFAR is shown to be a suboptimal one-class classifier: a Euclidian distance classifier and a quadratic discriminant with a missing term for one-parameter and two-parameter CFAR, respectively. We make a contribution toward enabling an objective design and implementation for target detection in SAR imagery.

• The paper presents a taxonomy categorizing detection methods into single-feature, multifeature, and expert-system approaches for SAR imagery.
• The paper demonstrates CFAR's dual role as both a signal processing filter and a statistical one-class classifier, highlighting its limitations and optimization potential.
• The paper compares detection techniques under varied clutter conditions, offering actionable insights for advancing SAR automatic target recognition systems.

Overview of "Target Detection in Synthetic Aperture Radar Imagery: A State-of-the-Art Survey"

The paper entitled "Target Detection in Synthetic Aperture Radar Imagery: A State-of-the-Art Survey" by Khalid El-Darymli, Peter McGuire, Desmond Power, and Cecilia Moloney presents a comprehensive review of methodologies pertinent to target detection in Synthetic Aperture Radar (SAR) imagery. Recognizing the pivotal role that target detection plays in an Automatic Target Recognition system for SAR imagery (SAR-ATR), the paper provides a structured taxonomy of detection methods and explores the theoretical underpinnings and practical implications of these techniques.

Taxonomy of Detection Methods

The paper posits a taxonomy centered around three primary approaches: single-feature-based, multifeature-based, and expert-system-oriented methods.

1. Single-Feature-Based Methods: These methods primarily rely on a single feature such as radar cross-section (RCS) to differentiate targets from the background. The Constant False Alarm Rate (CFAR) technique figures prominently within this category, with variants including CA-CFAR, SOCA-CFAR, GOCA-CFAR, and OS-CFAR. A key insight offered is that from a statistical pattern recognition perspective, CFAR can be considered a suboptimal one-class classifier.
2. Multifeature-Based Methods: This approach fuses multiple features extracted from the SAR imagery to refine detection performance. Techniques here may include multiresolution analysis using wavelet transforms or the inclusion of features like fractal dimensions.
3. Expert-System-Oriented Methods: These leverage artificial intelligence to integrate prior knowledge, such as scene segmentation and historical data, offering robust detection capabilities in SAR imagery. This methodology highlights the interplay between contextual information and classical detection algorithms to enhance detection efficiency.

Numerical Analysis and Comparative Insights

The paper compares different methodologies, employing a tabular format to provide a clear juxtaposition in terms of their operational contexts, such as feature types, clutter models, and applicability scenarios. For instance, it outlines the enhanced performance of multifeature-based methods in dealing with heterogeneous clutter scenarios which are challenging for single-feature approaches.

Novel Perspectives on CFAR Detection

A notable contribution of the paper is its dual-perspective analysis of CFAR detection from signal processing and statistical pattern recognition viewpoints:

• Signal Processing Perspective: CFAR is described as a finite impulse response band-pass filter, suggesting a convolutional approach to filtering SAR imagery in either spatial or frequency domains to extract target regions efficiently.
• Pattern Recognition Perspective: The paper explores how CFAR as an anomaly detector aligns with one-class classifiers, providing theoretical insights into CFAR's limitations and potential optimization pathways.

Implications and Future Directions

Practically, the paper underscores the importance of selecting appropriate probabilistic models for clutter representation, urging validation through goodness-of-fit tests tailored to SAR data characteristics. Theoretically, it emphasizes opportunities for advancements in CFAR methodologies by leveraging its interpreted link to discriminant analysis.

Looking ahead, the paper anticipates further research to refine context acquisition in SAR systems, notably for systems like Radarsat-2 under Spotlight operational modes. Moreover, extending this survey to address the classifier stages in the SAR-ATR processing chain is envisioned, hinting at potentials for enhanced end-to-end SAR imagery processing solutions.

In conclusion, this survey not only sheds light on the state-of-the-art techniques in SAR target detection but also lays a groundwork for future explorations into more sophisticated and contextually aware SAR-ATR systems.