The potential use of the interferometric coherence measured with Sentinel-1 satellites as input feature for crop classification is explored in this study. A one-year time series of Sentinel-1 images acquired over an agricultural area in Spain, in which 17 crop species are present, is exploited for this purpose. Different options regarding temporal baselines, polarisation, and combination with radiometric data (backscattering coefficient) are analysed. Results show that both radiometric and interferometric features provide notable classification accuracy when used individually (overall accuracy lies between 70% and 80%). It is found that the shortest temporal baseline coherences (6 days) and the use of all available intensity images perform best, hence proving the advantage of the 6-day revisit time provided by the Sentinel-1 constellation with respect to longer revisit times. It is also shown that dual-pol data always provide better classification results than single-pol ones. More importantly, when both coherence and backscattering coefficient are jointly used, a significant increase of accuracy is obtained (greater than 7% in overall accuracy). Individual accuracies of all crop types are increased, and an overall accuracy above 86% is reached. This proves that both features provide complementary information, and that the combination of interferometric and radiometric radar data constitute a solid information source for this application.

The application of satellite single-pass interferometric data to crop-type mapping is demonstrated for the first time in this work. A set of nine TanDEM-X dual-pol pairs of images acquired during its science phase, from June to August 2015, is exploited for this purpose. An agricultural site located in Sevilla (Spain), composed of fields of 13 different crop species, is employed for validation. Sets of input features formed by polarimetric and interferometric observables are tested for crop classification, including single-pass coherence and repeat-pass coherence formed by consecutive images. The backscattering coefficient at HH and VV channels and the correlation between channels form the set of polarimetric features employed as a reference set upon which the added value of interferometric coherence is evaluated. The inclusion of single-pass coherence as feature improves by 2% the overall accuracy (OA) with respect to the reference case, reaching 92%. More importantly, in single-pol configurations OA increases by 10% for the HH channel and by 8% for the VV channel, reaching 87% and 88%, respectively. Repeat-pass coherence also improves the classification performance, but with final scores slightly worse than with single-pass coherence. However, it improves the individual performance of the backscattering coefficient by 6–7%. Furthermore, in products evaluated at field level the dual-pol repeat-pass coherence features provide the same score as single-pass coherence features (overall accuracy above 94%). Consequently, the contribution of interferometry, both single-pass and repeat-pass, to crop-type mapping is proved.

Phase quality is a key element in the analysis of the deformation of the Earth's surface carried out with differential synthetic aperture radar interferometry. Various decorrelation sources may degrade the surface deformation estimates, and thus, phase filters are needed for this kind of application. The well-known Goldstein filter is the most widely used due to its simple implementation and computational efficiency. In the past years, improved filters have been proposed, which are based on this filter but introduce variations in the data processing. The effectiveness of these filters mostly depends on the size of the filtering window, the weight of the smoothed spectrum, and the kernel used to filter the spectrum. In this paper, we evaluate the performance of four of these filters and present a new method that outperforms all of them. The proposed filter is based on an iterative method in which the original phase is denoised progressively with adaptive filtering windows of different sizes. The effectiveness of the filter is controlled by the interferometric coherence, a direct indicator of the phase quality. Moreover, we introduce some modifications regarding the processing of the power spectrum. Specifically, we propose to smooth the original phase using a new filter which is based on a Chebyshev interpolation scheme. The performance of the new filter has been tested on both simulated and real interferograms, acquired by RADARSAT-2 and the Uninhabited Aerial Vehicle Synthetic Aperture Radar, which mapped two different geological events that caused surface deformation.

InSAR coherence is a promising parameter for land-cover classification and mapping. The ESA SEOM SInCohMap project is devised to test and analyze multi-temporal InSAR coherence potentialities exploiting dense multitemporal data from the Sentinel-1 constellation. In the framework of the project, this paper shows the first classification results using machine learning algorithms over a two-year period of InSAR coherence data. The evaluation is performed on the test site of Doñana (Seville, Southwestern Spain), mainly an agricultural area where different land covers can be identified. Classification results exploiting InSAR coherence shows accuracies around 80 % for this site.

Speckle suppression in synthetic aperture radar (SAR) images is a key processing step which continues to be a research topic. A wide variety of methods, using either spatially-based approaches or transform-based strategies, have been developed and have shown to provide outstanding results. However, recent advances in deep learning techniques and their application to SAR image despeckling have been demonstrated to offer state-of-the-art results. Unfortunately, they have been mostly applied to single-polarimetric images. The extension of a deep learning-based approach for speckle removal to polarimetric SAR (PolSAR) images is complicated because of the complex nature of the covariance matrices measured for every image pixel, the properties of which must be preserved during filtering. In this work, we propose a complete framework to remove speckle in polarimetric SAR images using a convolutional neural network. The methodology includes a reversible transformation of the original complex covariance matrix to obtain a set of real-valued intensity bands which are fed to the neural network. In addition, the proposed method includes a change detection strategy to avoid the neural network to learn erroneous features in areas affected by temporal changes, so that the network only learns the underlying speckle component present in the data. The method is implemented and tested with dual-polarimetric images acquired by Sentinel-1. The different experiments carried out in this work show that the proposed methodology provides good results in both speckle reduction and resolution preservation, similar to or even better than the ones obtained with state-of-the-art methods. Also importantly, it is shown that the neural network is not generating artifacts or introducing bias in the filtered images, making them suitable for further polarimetric processing and exploitation.

A geometrical decorrelation constitutes one of the sources of noise present in Synthetic Aperture Radar (SAR) interferograms. It comes from the different incidence angles of the two images used to form the interferograms, which cause a spectral (frequency) shift between them. A geometrical decorrelation must be compensated by a specific filtering technique known as range filtering, the goal of which is to estimate this spectral displacement and retain only the common parts of the images’ spectra, reducing the noise and improving the quality of the interferograms. Multiple range filters have been proposed in the literature. The most widely used methods are an adaptive filter approach, which estimates the spectral shift directly from the data; a method based on orbital information, which assumes a constant-slope (or flat) terrain; and slope-adaptive algorithms, which consider both orbital information and auxiliary topographic data. Their advantages and limitations are analyzed in this manuscript and, additionally, a new, more refined approach is proposed. Its goal is to enhance the filtering process by automatically adapting the filter to all types of surface variations using a multi-scale strategy. A pair of RADARSAT-2 images that mapped the mountainous area around the Etna volcano (Italy) are used for the study. The results show that filtering accuracy is improved with the new method including the steepest areas and vegetation-covered regions in which the performance of the original methods is limited.