In remote sensing, polarimetric synthetic aperture radar (PolSAR) data provide rich information for land cover characterization, particularly for vegetation monitoring. In this study, we investigate the statistical behavior of two roll-invariant geodesic parameters, alpha GD and PGD, derived from PolSAR data for crop characterization. These parameters are computed from geodesic distances in the space of 4 & times; 4 real matrices. We analyze their temporal evolution across different crop types and growth stages. Using the generalized additive model for location, scale, and shape (GAMLSS) framework, we model the relationships among alpha GD, PGD, crop type, and growth stage, and propose a classifier based on their joint stochastic information. Results from a comparative study with three existing methods show modest overall accuracies; however, the proposed approach consistently performs better in scenarios with less distinct class boundaries. Our results demonstrate that the GAMLSS approach effectively distinguishes between different crop types and their growth stages. Furthermore, the combined use of both parameters yields stronger discriminative power than using either parameter alone, thereby improving the characterization of crop types and phenological stages.

Synthetic aperture radar (SAR) change detection is a key tool for monitoring land-use and land-cover dynamics under all-weather and day/night conditions, yet its performance is often hindered by speckle noise, weak exploitation of spatial context, and limited modeling of nonlinear temporal dependencies. These issues make it difficult to reliably separate change and no-change regions, particularly in heterogeneous and transitional areas. This article presents an enriched SAR change detection via copula regression technique, which adopts a nonparameteric framework designed to address these challenges through dependence-aware and spatially consistent modeling. Particularly, the core contribution lies in the use of local copula density features, which capture nonlinear temporal dependencies between bitemporal SAR observations in a distribution-free manner and provide strong discriminative power under noise. To further stabilize these dependencies, spatial weighting kernels are employed to enhance both the difference image and the input images while preserving spatial structure. Finally, an enriched feature construction strategy is proposed to explicitly encode neighborhood-level information, by which coherent change maps can be produced using a regularized kernel logistic regression model with spatial consistency constraints. Experimental results on three diverse SAR datasets demonstrate the effectiveness of the proposed framework, achieving the highest Kappa scores on two datasets with relative Kappa improvement of up to 1.85% (up to 13.77% statistical gain) over the best baseline on San Francisco and/or Sulzberger datasets, while remaining competitive with recent deep learning approaches on the more challenging Ottawa dataset.

Detecting departures from the fully-developed speckle hypothesis in Synthetic Aperture Radar (SAR) imagery is essential for reliable interpretation in remote sensing applications. To address this challenge, this paper proposes a statistical test based on a nonparametric estimator of Tsallis entropy, a non-additive generalization of Shannon entropy that provides enhanced sensitivity to heavy-tailed distributions characteristic of textured SAR data. The estimator incorporates bootstrap correction to improve accuracy with small sample sizes. The test is integrated within an adaptive windowing strategy that locally selects the optimal window size based on regional homogeneity: larger windows in homogeneous regions to stabilize estimation and smaller windows in heterogeneous areas to preserve structural details. This combination yields an unsupervised statistical framework that generates per-pixel p-value maps for detecting departures from fully-developed speckle. Experimental validation using both simulated data and SAR imagery confirms the method’s precision in detecting texture variability and discriminating between homogeneous and heterogeneous regions. The proposed approach offers an interpretable tool for automated SAR image analysis without requiring training data or explicit parametric texture estimation.

Target characterization parameters are crucial for accurately capturing the physical and geometric properties of land cover targets. While numerous methods and descriptors have been proposed for full-polarimetric (full-pol) SAR data to enable detailed target characterization, dual-polarimetric (dual-pol) synthetic aperture radar (SAR)-based analyses have traditionally relied on backscatter intensities and conventional dual-pol descriptors. However, these conventional descriptors lack discrimination between elementary targets, such as trihedral (or surface) and dihedral scatterers, leading to ambiguities in characterizing diverse land cover targets. Recent studies introduced advanced dual-pol descriptors that effectively characterize "dihedral-like" and "surface-like" scattering behaviors in dualpol SAR data. In particular, a dual-pol target characteristic parameter, alpha(-)((S)), was proposed and subsequently used to derive the dual-pol three-component scattering powers, decomposing the total power into relatively coarse scattering components. Further developments, combined alpha(-)((S)) with wave entropy, H-w, to partition the H-w-alpha(-)((S)) plane into eight cluster zones, each representing specific scattering characteristics. This study proposes a zone-based scattering power (ZBP) components that exploit the centroids and areas of the cluster zones together with a metric associated with the target's position in the H-w-alpha(-)((S)) plane to derive extended scattering power components from dual-pol SAR data. The proposed framework decomposes the total power into finer scattering components. Assessment over diverse land cover targets demonstrates that the proposed ZBP components capture subtle variations in target scattering behavior and provide additional discriminatory information for enhanced target characterization.

Synthetic aperture radar (SAR) is a central tool for mapping scenarios on the Earth’s surface. Speckle, which is inherent to every image obtained with coherent illumination, significantly degrades the perceived SAR image quality, making it challenging to analyze and interpret. Thus, the choice of statistical models able to describe speckled data is fundamental. In this sense, the multiplicative approach has received considerable attention. A particular important case from this class is the G model proposed by Frery et al. [IEEE Transactions on Geoscience and Remote Sensing, vol. 35, no. 11, pp. 648–659, 1997], which encompasses several relevant distributions. In this paper, we extend such class, proposing models that enhance classical distributions from the G family, such as the G0 and K distributions. We obtain their statistical properties such as moments, maximum likelihood estimators, Mellin-based log-cumulants (LCs), and their covariance matrices. Applications to actual data provide evidence that the new distributions outperform usual G models.

Monitoring and detecting changes in sensor and ubiquitous network data streams (data drifts) is essential to ensure reliability, adaptability, and security in dynamic environments. We present Quantile-based Change Voting with Entropy and Complexity (QCV-HxC), an unsupervised framework for drift detection based on the temporal evolution of permutation entropy and statistical complexity, both derived from ordinal patterns representations. This framework combines smoothed temporal derivatives of these measures using composite functions, identifies change points using quantile-based thresholds, and then applies a voting mechanism across detection windows. We evaluate QCV-HxC on both real-world data with temporal dependencies (Electricity Market) and synthetic streams exhibiting abrupt concept shifts (RTG_2abrupt). Experimental results demonstrate that QCV-HxC achieves an average $\mathbf {F_{1}}$ -score of 0.80, surpassing traditional detectors such as ADWIN and Page-Hinkley-while operating entirely without label supervision. These findings indicate that the proposed framework provides a robust, efficient, and interpretable solution for real-time drift detection and online monitoring in wireless, sensor, and ubiquitous computing environments.

SAR interferometry (InSAR) provides a framework for extracting high-resolution topographic information and detecting surface deformation. By analyzing the phase difference between radar acquisitions obtained at different times, one can characterize landscape geometry and surface changes. However, inherent phase noise often compromises the reliability of the resulting interferometric products. Consequently, there is a sustained need for spatial filtering techniques that suppress noise while preserving structural integrity and resolution. This work addresses the challenge of filtering the unwrapped phase, a process traditionally reliant on accurate coherence images to identify reliable pixels. We evaluate three statistically based spatial filters for phase noise reduction. The Enhanced Lee filter, which utilizes spatial adaptation and a physically grounded probability model, serves as the baseline for comparison. We examine the Gierull model, which improves computational efficiency by restricting the parameter space. To further reduce execution time, we propose and evaluate two empirical alternatives: the truncated wrapped normal (TcN) and the truncated wrapped Cauchy (TcC) distributions. Results indicate that these empirical models significantly reduce computational demand without degrading the quality of the filtered phase. We assess performance using a simulated dataset for objective validation alongside InSAR imagery of La Cumbre volcano, Los Alamos, and Robledo volcano. While the proposed models demonstrate significant gains in computational efficiency compared to current methods, we identify numerical integration as a primary bottleneck in the filtering process; this challenge warrants further investigation. Our results indicate that empirical statistical models provide a viable path for accelerated InSAR processing with accuracy equivalent to traditional, computationally intensive approaches.

O monitoramento de tráfego de rede é essencial para compreender o comportamento da infraestrutura e avaliar a integridade de seus componentes. O aprendizado federado tem se destacado como uma abordagem promissora para sistemas de defesa baseados nesse monitoramento, permitindo o treinamento distribuído de modelos sem compartilhamento direto de dados. No entanto, métodos tradicionais assumem um ambiente federado composto apenas por clientes honestos, ignorando a possibilidade de ataques de envenenamento de rótulos (label poisoning). Este trabalho propõe um novo arcabouço de aprendizado federado robusto contra ataques de rede, com foco na mitigação de clientes maliciosos. Nossa abordagem emprega técnicas de Redes Siamesas para quantificar a aderência dos dados e ajustar dinamicamente a ponderação das contribuições de cada cliente, fortalecendo a resiliência do modelo contra manipulações adversárias. Os resultados mostram que nossa estratégia não apenas melhora a detecção de ataques, mas também reduz significativamente o impacto de envenenamento de rótulos no aprendizado federado.

Polarimetric Synthetic Aperture Radar (PolSAR) sensors have emerged as a groundbreaking remote sensing technology. They enable the acquisition of the amplitude, phase, and orientation of electromagnetic waves across multiple polarizations. This capability provides enhanced potential for detailed environmental analysis. However, challenges such as complex data structures, non-Gaussian noise properties, and low signal-to-noise ratios pose significant barriers to the effective use of PolSAR data. Existing methods for modeling and analyzing PolSAR data are predominantly parametric and rely on assumptions that may fail under certain conditions. Aware of these limitations, this study introduces the use of U-statistics for PolSAR data analysis. Using information from the diagonal intensities of the covariance matrix, we propose a hypothesis testing mechanism to assess sample homogeneity a top-down hierarchical separability analysis, and a U-statistics-based classification approach. The proposed procedures are validated using an ALOS PALSAR image of the Amazonian region. The results show the robustness and effectiveness of the proposed methods, offering a reliable framework for analyzing and classifying PolSAR data under non-parametric assumptions.

This study introduces two adaptive scattering descriptors for analyzing SAR data: (1) Radar Vegetation Index and (2) Scattering Complexity. The term “adaptive” highlights their flexibility, as these methods can be applied seamlessly to full-, dual-, and compact-polarimetric SAR data, ensuring suitability for diverse applications. These descriptors are evaluated using temporal Radarsat-2 data to analyze the growth stages of paddy in Vijayawada, India. The descriptors reveal interesting behaviors in the crop's growth dynamics, capturing key variations in scattering properties and vegetation structure at different stages.

Change detection in synthetic aperture radar imagery is challenging due to noise, complex nonlinear features, and variations in underlying data distributions. Traditional clustering-based methods often make strong assumptions about data structure, struggle with noise, and fail to handle the intricate characteristics of radar images effectively. Furthermore, deriving change maps directly from the difference image can introduce confusion and increase detection errors. This article presents a novel change detection approach that employs the hierarchical density-based clustering algorithm to improve the accuracy and robustness of change map generation. The method automatically identifies clusters of varying densities while filtering out noise, ensuring a more precise change detection process. A new hyperparameter optimization strategy is introduced to enhance clustering performance by tuning key parameters based on the silhouette index. In addition, an adaptive thresholding mechanism leverages cluster probability and statistical measures, including location and dispersion, to refine the selection of meaningful changes. The methodology consists of computing a difference image using a logarithmic ratio operator, optimizing clustering parameters, computing high-change probability clusters, and generating a refined change map. Experimental results on multiple datasets demonstrate the effectiveness of the proposed approach in adapting to different data complexities, improving detection accuracy, and minimizing false positives compared to existing methods.

Quantifying heterogeneity in synthetic aperture radar (SAR) data is critical for accurate geophysical interpretation and remote sensing applications. We propose a test statistic based on a nonparametric estimation of R & eacute;nyi entropy to characterize return heterogeneity from SAR intensity data. The statistic is refined using bootstrap to improve its stability, size, and power. This approach enhances heterogeneity quantification by capturing scale-dependent variations and addressing data-driven uncertainty. Experimental results establish the robustness of the proposed method in distinguishing heterogeneity patterns.

We devise a constant false alarm rate (CFAR) detector for detecting targets in clutter regions with varying degrees of heterogeneity. The CFAR detector employs the G(0) distribution for characterizing the statistics of background Synthetic Aperture Radar (SAR) clutter amplitude data. We express the Method of Log-Cumulants (MoLC) estimator for the shape parameter of G(0) -distribution in closed-form using an approximation of the polygamma function. We quantify the effectiveness of employing a polygamma approximation in MoLC estimation for clutter regions with varying levels of heterogeneity using a Monte Carlo simulation. Exploiting the dependencies between hypergeometric functions and the Fisher-Snedecor distribution function, we formulate a closed-form solution for the CFAR detection threshold. Furthermore, we derive a closed-form expression of the detection probability for the proposed CFAR detector in terms of clutter model parameters and the signal-to-clutter ratio (SCR). We present analytical results that confirm the efficacy of the CFAR detector in various clutter environments, as well as for different SCR levels, compared to the CFAR-WBL, CFAR-LGN, CFAR- $\mathcal {K}$ , and existing CFAR- G(0) detectors. We further assess the performance of the CFAR detector on ALOSPALSAR and MSTAR (Moving and Stationary Target Acquisition and Recognition) data, which represent sea clutter and vegetation clutter, respectively. The proposed detector achieves accurate detection results compared to the aforementioned state-of-the-art CFAR detectors. Finally, experimental results illustrate the computational effectiveness of the proposed CFAR- G(0 ) detector compared to CFAR- $\mathcal {K}$ and conventional CFAR- G(0) detectors.

In this letter, considering the effectiveness of 2-D principal component analysis (2DPCA) on the exploration of local spatial relationships, a reconstruction-based 2DPCA (Rec-2DPCA) operation was designed for feature extraction and injected into the architecture of PCANet for change detection of bitemporal synthetic aperture radar (SAR) image. Specifically, as the projection of an image patch on one eigenvector computed by 2DPCA breaks the one-to-one relationship between feature map and eigenvalue, we adopted Rec-2DPCA at various network layers and developed two variants of PCANet, namely, 2DPCANet and (2-D + 1-D)PCANet. In the experiments, using three real SAR image datasets, we analyzed the performance of all comparison methods, and our proposals achieved a more appealing performance than other methods.

The cald & eacute;n (Neltuma caldenia) forest, a xerophytic low-stature ecosystem in central Argentina, faces increasing threats from land use change and desertification. This study assesses the capability of full-polarimetric L-band SAR data from the Argentine SAOCOM-1A satellite to characterise forest attributes in this ecosystem. We computed the Generalised Radar Vegetation Index (GRVI) and compared it with aboveground biomass and tree canopy cover data from the Second National Forest Inventory, under fire and non-fire conditions. We also assessed other SAR indices and polarimetric decompositions. GRVI values exhibited limited variability relative to the broad range of field-estimated biomass, and most regression models were not statistically significant. Nevertheless, GRVI effectively distinguished woody from non-woody vegetation and showed a weak correlation with canopy cover. Statistically significant, albeit weak, correlations were also observed between biomass and specific polarimetric components, such as the helix term of the Yamaguchi decomposition and the Pauli volume component. Key challenges included limited spatial and temporal coverage of SAOCOM-1A data and the distribution of field plots. Despite these limitations, our results support the use of GRVI for land cover monitoring in semiarid regions, emphasising the importance of multitemporal data, integration with C-band SAR, and enhanced field sampling to improve forest attribute modelling.

Optical sensors are not able to capture images under the effect of severe weather conditions and poor lighting, making the task of extracting information from optical images difficult. The application of convolutional neural networks (CNNs) to Synthetic Aperture Radar (SAR)/ Polarimetric Synthetic Aperture Radar (PolSAR) images can mitigate these conceptual limitations characteristic of optical images due to the electromagnetic nature of SAR / PolSAR data capture sensors.The problem we propose in the research is finding a suitable configuration for the CNN to detect roads in SAR / PolSAR images with accuracy comparable to or better than the state-of-the-art and with excellent computational efficiency.

SAR Data are affected by speckle, a non-additive and non-gaussian interference noise-like pattern. The distribution these data follow is paramount for their processing and analysis. Good statistical models provide flexibility and accuracy, often at the cost of using several parameters. The $\mathcal{G}^0$ distribution is one of the most successful models for SAR data. It includes the Gamma law as a particular case which arises in the presence of fully developed speckle. Although the latter is a limit distribution of the former, using the same estimation technique for the more general model is numerically unfeasible. We developed a test statistic based on an bootstrap-improved non-parametric estimator of the Shannon entropy for assessing departures from the fully-developed speckle hypothesis. We show the adequacy of the proposal with simulated and SAR data.

Synthetic aperture radar (SAR) is an efficient and widely used remote sensing tool. However, data extracted from SAR images are contaminated with speckle, which precludes the application of techniques based on the assumption of additive and normally distributed noise. One of the most successful approaches to describing such data is the multiplicative model, where intensities can follow a variety of distributions with positive support. The GI0 model is among the most successful ones. Although several estimation methods for the GI0 parameters have been proposed, there is no work exploring a regression structure for this model. Such a structure could allow us to infer unobserved values from available ones. In this work, we propose a GI0 regression model and use it to describe the influence of intensities from other polarimetric channels. We derive some theoretical properties for the new model: Fisher information matrix, residual measures, and influential tools. Maximum likelihood point and interval estimation methods are proposed and evaluated by Monte Carlo experiments. Results from simulated and actual data show that the new model can be helpful for SAR image analysis.

Speckle noise is a phenomenon that occurs in the acquiring image process with coherent illumination, such as synthetic aperture radar, ultrasound, or laser systems. It is hard to eliminate due to its stochastic and multiplicative nature. In this work, a despeckling filter technique is proposed based on modeling the data using the G(I)(0) distribution and applying a hypothesis test built from the Tsallis entropy. The present proposal is applied to synthetic and real images and, in both cases, the efficiency in removing speckle noise is evaluated using standard metrics. Furthermore, its quality is compared with the FANS filter, obtaining encouraging results.

This column is meant to provide information about an exciting opportunity offered by the IEEE Geoscience and Remote Sensing Society (GRSS) to its members in conjunction with IEEE Press.