In this paper we consider physics-informed detection of terrain material change in radar imagery (e.g., shifts in permittivity, roughness or moisture). We propose a lightweight electromagnetic (EM) forward model to simulate bi-temporal single-look complex (SLC) images from labelled material maps. On these data, we derive physics-aware feature stacks that include interferometric coherence, and evaluate unsupervised detectors: Reed-Xiaoli (RX)/Local-RX with robust scatter (Tyler's M-estimator), Coherent Change Detection (CCD), and a compact convolutional auto-encoder. Monte Carlo experiments sweep dielectric/roughness/moisture changes, number of looks and clutter regimes (gamma vs K-family) at fixed probability of false alarm. Results on synthetic but physically grounded scenes show that coherence and robust covariance markedly improve anomaly detection of material changes; a simple score-level fusion achieves the best F1 in heavy-tailed clutter.
Detecting subtle, spatially localised changes in terrain materials from radar imagery underpins applications in infrastructure monitoring, environmental surveillance and security [1,2,3]. In coherent radar, long-standing strategies for detecting changes span intensitybased methods and coherent change detection (CCD), which exploits interferometric coherence estimated between phase-registered image pairs with material or structural changes reducing the coherence magnitude [4,5,6]. While state-of-the-art deep learning models have advanced bi-temporal change mapping, most works target pixel stacks statistically, with limited explicit ties to the EM material properties that actually derive radar backscatter and decorrelation [7,8].
The radar backscatter obtained from terrain depends on a number of factors, such as the complex permittivity and surface roughness parameters of the terrain, as well as the radar geometry (incidence angle). Widely used rough-surface models such as the Integral Equation Model (IEM/AIEM) provide validated relationships between these parameters and the radar backscatter [2,9], whereas, popular anomaly detectors like the Reed-Xiaoli (RX) score assume a (often Gaussian) background distribution and flag departures from this via Mahalanobis distance [10]. In radar imagery, heavy-tailed clutter is common, so a fixed global threshold does not preserve a stable false alarm rate across heterogenous backgrounds; using robust covariance estimators (e.g., Tyler’s M-estimator) makes detectors more CFAR-like, i.e., the decision threshold adapts to local clutter statistics so the false alarm probability remains approximately constant across the scene [3,11].
Despite the different approaches to anomaly detection in radar imagery, there remains a lack of reproducible, controlled studies that link material-level changes (e.g., moisture-driven permittivity or roughness changes) to expected changes in backscatter and coherence, and the relative performance of RX/robust-RX, CCD and modern unsupervised models. Material change (e.g., dry→wet soil, asphalt ageing, vegetation removal) alters permittivity and roughness, shifting Fresnel/rough-surface reflectivity and speckle statistics; simultaneously, coherence decreases in changed regions. Thus, coherence-aware features and robust background modelling are well-matched not only to radiometric change but also to material change [2,3,5].
In this work, we address the problem of terrain material change by proposing a physics-informed anomaly detection framework. We use a compact, IEM-inspired forward model to map labelled material maps to bi-temporal single look complex (SLC) images. Perpixel parameters for permittivity, rms height, correlation length and incidence angle determine mean backscatter; multiplicative speckle and controlled cross-epoch correlation generate SLC pairs that emulate decorrelation for genuine material/structural change [2,9]. From these images, a physics-aware feature stack is estimated in local windows combining log-intensities, simple texture, incidence angle, and interferometric coherence γ [5]. We then perform a comparative study of unsupervised detectors to highlight regimes where physicsaware features and robust covariance are decisive. We consider both global and local RX with a robust scatter estimator for heavy-tailed clutter [3,10,11]; CCD via decorrelation [6]; and a lightweight convolutional auto-encoder trained on unchanged tiles [7,12]. The remainder of this paper is structured as follows: Section 2 details the proposed forward model and Section 3 describes the features and detectors. In Section 4 we outline the datasets used and the Monte-Carlo protocol which covers changes in permittivity, rms height, number of looks and Signal to Noise Ratio (SNR). Section 5 presents results and ablation data reporting Receiver Operating Characteristic-Area Under the Curve (ROC-AUC), Average Precision (AP) and F1 score at low Probability of False Alarm (PFA). Section 6 concludes the paper, highlighting that on synthetic, but physically grounded data, coherence-aware methods decisively outperform traditional intensity-based detectors, with a simple score-level fusion achieving the highest performance.
In this section, we describe our SLC image formation framework using a physics-informed forward model. In our model we assume a monostatic, side-looking geometry with a known radar pose. A base digital elevation model (DEM) provides local surface normal n(r) (with r the geospatial location in the DEM/frame); the local incidence angle is θ(r) = arccos( k • n), where k is a unit vector denoting the look direction. A labelled material map assigns, per pixel p, complex permittivity εr,p(f ), where f is the radar operating frequency and roughness parameters: rms height σp and correlation length lc,p. These parameters control microwave backscatter according to rough-surface EM models such as IEM/AIEM [2,9,13,14].
The exact IEM/AIEM backscatter r
This content is AI-processed based on open access ArXiv data.