Polarimetric decompositions are used to separate scatterers and identify their physical parameters by analyzing backscattering, coherence, or covariance matrices. Each cell within polarimetric SAR data is seen as a coherent or incoherent combination of different scattering mechanisms. However, targets are not perfectly characterized by these matrices due to the presence of noise components. The main objective of this study is to remedy the latest issue through proper noise effect elimination. Hence, we propose the re-estimation of the coherence matrix, by incorporating a processing phase that searches for the number of elementary scattering mechanisms in each cell. This first step is based on the eigenvalues, which exploit the advantage of polarization basis independent of the eigenvectors. In the second step, a reduced space is defined by the eigenvectors selected, according to the cases of the first step, as those contributing to the construction of the target, excluding those judged to contribute to noise. The characteristic vector and/or the coherence matrix of the average target is then reconstructed in this new space in three different ways: summation of the elementary coherence matrices, applying Bernoulli's probability law, and orthogonal projection on the reduced space. Finally, the Freeman Durden polarimetric decomposition and the H- $\overline{\alpha }$ Wishart classification are used to show the effectiveness of the process in terms of dominant scattering mechanism identification. Their application on simulated data and on fully-polarized RadarSat-2 images of the city of Algiers attests to the performance of the proposed methodology to improve the identification of dominant scattering mechanisms.
24 October 2024 in IEEE Geoscience and Remote Sensing Letters
The detection of multiple scatterers within each resolution cell is an open research subject in Synthetic Aperture Radar tomography. For over a decade, the Generalised Likelihood Ratio Test (GLRT) detector has been implemented along with its variants, allowing the generation of height maps and 3D point clouds with good precision. However, they are limited by the grid search during the optimization of the Maximum Likelihood function. In order to mitigate this, we propose a gridless version of GLRT where the Particle Swarm Optimization (PSO) method is used to locate the minima. The conducted analysis of the proposed detector with respect to the state-of-the-art methods behavior on simulated and real datasets proved the effectiveness of PSO-GLRT in terms of height accuracy and computational cost. The evaluation metrics Root Mean Square Error, Accuracy, and Completeness have been used as a quantitative improvement indicator for estimated height assessment.
Height estimation of scatterers in complex environments via the Tomographic Synthetic Aperture Radar (TomoSAR) technique is still a valuable research field. The parametric spectral estimation approach constitutes a powerful tool to identify the superimposed scatterers with different complex reflectivities, located at different heights in the same range–azimuth resolution cell. Unfortunately, this approach requires prior knowledge about the number of scatterers for each pixel, which is not possible in practical situations. In this paper, we propose a method that analyzes the scree plot, generated from the spectral decomposition of the multidimensional covariance matrix, in order to estimate automatically the number of scatterers for each resolution cell. In this context, a properly improved regularization step is included during the reconstruction process, transforming the parametric MUSIC estimator into a non-parametric method. The experimental results on two data sets covering high elevation towers, with different facade coating characteristics, acquired by the TerraSAR-X satellite highlighted the effectiveness of the proposed regularized MUSIC for the reconstruction of high man-made structures compared with classical approaches.
9 January 2023 in IEEE Transactions on Geoscience and Remote Sensing
Synthetic Aperture Radar tomography has recently received particular interest from the remote sensing community, due to its ability to provide 3D reconstructions of environments with complex structures. Unfortunately, different forms of decorrelations and processing errors affect the quality of the resulting 2D / 3D images. One way to cope with the impact of these nuisances is to apply an appropriate filtering to the interferometric data stack as a preprocessing step. The first obstacle to be dealt with, especially in urban areas, is to define a filter whose parameters have to be set in such a way to improve smoothing capabilities while preserving edges. To this aim, the main objective of this paper is twofold: 1) application of a spatio-temporal contextual filter whose parameters depend on 3D quality indicators of the multi-baseline interferometric image stack, 2) evaluation of the denoising effect on the application of nonparametric spectral estimation and detection algorithms. For that, we consider several quantitative metrics to assess on one hand the filtering performances, and on the other hand, its impact on the reflectivity function recovered from conventional tomographic inversion and detection methods. Experimental results from a set of TerraSAR-X images highlight the efficiency of the filtering process by improving the scatterers’ detection and height localization, of a man-made structure.
The Tomographic Synthetic Aperture Radar (TomoSAR) technique is widely used for monitoring urban infrastructures, as it enables the mapping of individual scatterers across additional dimensions such as height (3D), thermal dilation (4D), and deformation velocity (5D). Retrieving this information is crucial for building management and maintenance. Nevertheless, accurately extracting it from TomoSAR data poses several challenges, particularly the presence of outliers due to uneven and limited baseline distributions. One way to address these issues is through statistical detection approaches such as the Generalized Likelihood Ratio Test, which ensures a Constant False Alarm Rate. While effective, these methods face two primary limitations: high computational complexity and the off-grid problem caused by the discretization of the search space. To overcome these drawbacks, we propose an approach that combines a quick initialization process using Fast-Sup GLRT with local descent optimization. This method operates directly in the continuous domain, bypassing the limitations of grid-based search while significantly reducing computational costs. Experiments conducted on both simulated and real datasets acquired with the TerraSAR-X satellite over the Spanish city of Barcelona demonstrate the ability of the proposed approach to maintain computational efficiency while improving scatterer localization accuracy in the third and fourth dimensions.
30 May 2024 in IEEE Geoscience and Remote Sensing Letters
The reconstruction of high-elevation natural and artificial structures through Synthetic Aperture Radar (SAR) tomography has been an active research topic owing to its significance in various earth science applications. However, the complexity of this task arises from inaccuracies in the estimated reconstruction, attributed to factors such as low signal-to-noise ratios, decorrelations, few and uneven measurements, and overlapping scatterers. The utilization of iterative spectral estimation methods has been demonstrated to be beneficial in addressing some of these inaccuracies. Thus, selecting the best method within this class constitutes a challenge. In this context, our letter aims to propose a generalized formula linking the maximum likelihood-based iterative methods via a regularization parameter. The behavior of the latter is analyzed for several values in order to unveil the potential of the proposed approach in achieving a balance between noise reduction and detection performance. The experimental study has been conducted on simulated and real SAR data acquired by airborne and spaceborne systems covering tropical forest and build-up areas. The obtained results show the effectiveness and performance of the optimal regularization parameter to eliminate noise while preserving scatterers’ contribution.