This work introduces a new methodology to estimate vegetation parameters from remote sensing data exploiting machine learning techniques. The workflow uses appropriate active learning and regression strategies, based on ensembles, to significantly reduce the number of calibration data needed to run such algorithms. Experimental results, concerning the estimation of vegetation biomass and nitrogen content, will show that it is possible to reduce the calibration dataset up to 80% with a degradation of the estimation performance, against literature data, in the order of 5% for biomass and 15% for nitrogen content.

Different anthropogenic activities, both legal and illegal, can potentially induce soil contamination, which represents a risk to both the environment and human health. Utilizing Remote Sensing technologies to identify the effects of soil contamination on the local plant enables rapid recognition of these environmental hazards, facilitating more effective management and control. In this context, the use of hyperspectral data acquired from UAVs can be very useful. Hyperspectral imaging captures a much larger number of narrow spectral bands than multispectral imaging, allowing for more precise and detailed spectral information to be extracted from the captured data. In this paper, we leverage a set of hyperspectral indices to train a Support Vector Machine (SVM) for discriminating between plants grown in soil contaminated with substances such as Pb, Cr, Zn, and Benzopyrene, and those grown in uncontaminated soil. This approach enables us to harness the rich spectral information provided by hyperspectral data to better identify vegetation changes due to contamination, enabling us to detect environmental issues with higher reliability compared to using multispectral data.

The identification of contaminated sites is a key application in remote sensing. Major challenges are typically related to the difficulties in the spectral characterization of the targets. The detection performance can be improved by exploiting a suitable spatial description of the scene. In this work, the use of spatial features based on the Scattering Transform for detection of micro-dumps and greenhouses is proposed. These features are jointly exploited with spectral ones to improve the classification accuracy within a multi -class classifier. The obtained results show the relevance of spatial features in the detection of the targets of interest.

Change detection is one of the most addressed topics in the remote sensing community. When performed on synthetic aperture radar images, the most critical issues are as follows: 1) the labeling of the identified changing patterns and 2) the scarce robustness of classic pixel-based approaches based on threshold segmentation of an appropriate change index, which tend to fail when multiple changes are present in the study area. In this work, a new methodology for unsupervised change detection in vegetation canopy is presented. It overcomes these limitations by exploiting multitemporal geographical object-based image analysis with the aim to make the intrinsic semantic of data emerge and direct the processing toward the identification of precise classes of changes through dictionary-based preclassification and fuzzy combination of class-specific information layers. The proposed methodology has been tested in ten different experiments covering agriculture and clear-cut deforestation applications. The results, validated against literature methods, highlighted the superiority of the proposed approach, which was quantitatively assessed in terms of standard classification quality parameters. On agriculture experiments, it allowed for an average increase in the detection accuracy of about 11% with respect to the best performing literature method, with an increment of the false alarm rate in the order of 0.5%. In case of deforestation, the registered detection accuracy was comparable to that achieved by the literature, while the most significant benefit was the reduction, of more than one-third, of the number of detected false deforestation patterns. Overall, the main characteristics of the proposed architecture are the robustness and the lack of any supervision, which makes it very well-suited for operational scenarios.

In the above article[1], the author affiliations were incorrectly listed. The correct affiliations are as follows:

In the event of an accidental oil spillage from buried pipelines, the leakage may be detected fairly late, sometimes months after it started, with economic and environmental impacts that are difficult to recover. An early detection is desirable but difficult to achieve when the pipelines are buried, hidden by thick vegetation or in quite isolated areas. If the extent of buried pipelines network is also considered the problem may appear cumbersome. Satellites may support an early detection and, a feasibility study proving so, is described in this paper. A controlled spillage exercise has been organized in the UK and satellite data acquisitions tasked. Synthetic Aperture Radar datasets have been acquired over the site, before and after the controlled spillage, in three different tests with different quantities of red diesel spilled. The design of the whole exercise is described in the paper in addition to the SAR data processing. Preliminary results show that a significant change in the mean backscattering coefficient (a decrease of about 1dB) is appreciated in X-band SAR data when the volume of oil, spilled or poured, reaches values in the order of 160 gallons.

Land cover mapping is one of the classic applications of synthetic aperture radar remote sensing. However, despite of the algorithmic progress in classification techniques, the semantic content of available maps does remain unchanged, with only a few macro-classes (like water, forest, urban, and bare soil) being discriminated in the majority of the works from past years. In this paper, a methodology to extract a higher level semantics from synthetic aperture radar images is presented. It is based on coupling pixel-based clustering with object-based image analysis and contextual information. Preliminary results have been produced from multitemporal SAR datasets over a forest area in Colombia. They demonstrate that the synergic exploitation of pixel and object information can provide higher quality land cover results and more information to map users.

The Urban Heat Island (UHI) effect is defined as an increase of the air and surface temperature inside a city compared to surrounding rural areas. This increment can be of several degrees, thus exposing populations to serious health risks, especially in hot developing countries, where the majority of the world's megacities are located. The UHI effect has been widely studied in the past with local methods employing field sensors. The use of satellites moved the analysis from local to city scale, but long-term investigations have been so far limited by storage and computational capacities. In this work, both ESA and NASA heritage data are used to study temporal evolution of the UHI of the city of Chennai, India over a 14-year period. The Google Earth Engine is exploited to process the available large dataset in a reasonable time. Results show that the UHI of Chennai has grown of 450% over time and that its main drivers are average temperature and city expansion.

In the Synthetic Aperture Radar (SAR) framework many detection algorithms and techniques have been published in the recent literature; however the detection of vessels whose dimensions are in the order of the image spatial resolution is still challenging in rough sea state scenarios. This issue is addressed in the paper presented here by comparing rationale and performance of two detectors developed by the same authors: the Generalized Likelihood Ratio Test (GLRT) and the Intensity Dual-Polarization Ratio Anomaly Detector (iDPolRAD). Both detectors are tested on a dual-polarization VV/VH Interferometric Wide Swath Sentinel-1 image acquired over the Suruga Bay on the Pacific Coast of Japan. The theory is presented here and the two detectors are compared against the Cell Average-Constant False Alarm Algorithm (CA-CFAR) showing both better performance than CFAR in terms of false alarms rejection.

In this work, we present a new architecture for the analysis multitemporal SAR data combining classic synthetic aperture radar processing and geographical object-based image analysis. The architecture exploits the characteristics of the recently introduced RGB products of the Level-1α and Level-1β families, employing self-organizing map clustering and object-based image analysis aiming at the definition of opportune layers measuring scattering and geometric properties of candidate objects to classify. The obtained results have been compared with those given by literature and turned out to provide high degree of accuracy and negligible false alarms. The discussion is supported by an example concerning small reservoir mapping in semi-arid environment.

In this paper, we present a novel method for mapping flooded areas exploiting Sentinel-1 ground range detected products. The work introduces two novelties. As first, the input products. In fact, as far we know, no applications using these products has been so far presented in literature. Secondly, a new unsupervised methodology, based on the usage of opportune layers combined in a fuzzy decision system, is presented. Experimental results, obtained both on the single SAR image and on a couple of acquisitions in a change detection framework showed that our method is able to outperform the most popular classification techniques in terms of standard assessment parameters.

We present a new methodology for rapid flood mapping exploiting Sentinel-1 synthetic aperture radar data. In particular, we propose the usage of ground range detected (GRD) images, i.e., preprocessed products made available by the European Space Agency, which can be quickly treated for information extraction through simple and poorly demanding algorithms. The proposed framework is based on two processing levels providing event maps with increasing resolution. The first level exploits classic co-occurrence texture measures combined with amplitude information in a fuzzy classification system avoiding the critical step of thresholding. The second level consists of a change-detection approach applied to the full resolution GRD product. The discussion is supported by several experiments demonstrating the potentiality of the proposed methodology, which is particularly oriented toward the end-user community.

In this paper, we present a new framework for the fusion, representation, and analysis of multitemporal synthetic aperture radar (SAR) data. It leads to the definition of a new class of products representing an intermediate level between the classic Level-1 and Level-2 products. The proposed Level-1β products are particularly oriented toward nonexpert users. In fact, their principal characteristics are the interpretability and the suitability to be processed with standard algorithms. The main innovation of this paper is the design of a suitable RGB representation of data aiming to enhance the information content of the time-series. The physical rationale of the products is presented through examples, in which we show their robustness with respect to sensor, acquisition mode, and geographic area. A discussion about the suitability of the proposed products with Sentinel-1 imagery is also provided, showing the full compatibility with data acquired by the new European Space Agency sensor. Finally, we propose two applications based on the use of Kohonen's self-organizing maps dealing with classification problems.

In this paper, we present a new framework for high-level processing of time series images, with particular reference to Sentinel-1 data. The proposed methodology has the goal of enhancing the interpretation of SAR imagery through the production of physical-based RGB composites, which are particularly suited for being easily interpreted by the human photo-interpreter, lowering the expertise level required for managing SAR data.

In this paper we propose a methodology devoted to exploit high resolution radars for monitoring water bodies in semi-arid countries. The proposed approach is based on appropriate registration, calibration and processing of SAR data, producing information ready to use by end-users. The obtained results were used to (i) estimate a relationship between surface and volume of water stored in reservoirs and (ii) validate a hydrological model that simulates the time evolution of water availability.

This paper presents the multitemporal adaptive processing (MAP3) framework for the treatment of multitemporal synthetic aperture radar (SAR) images. The framework is organized in three major activities dealing with calibration, adaptability, and representation. The processing chain has been designed looking at the simplicity, i.e., the minimization of the operations needed to obtain the products, and at the algorithms' availability in the literature. Innovation has been provided in the cross-calibration step, which is solved introducing the variable amplitude levels equalization (VALE) method, through which it is possible to establish a common metrics for the measurement of the amplitude levels exhibited by the images of the series. Representation issues are discussed with an application-based approach, supported by examples with regard to semiarid and temperate regions in which amplitude maps and interferometric coherence are combined in an original way.

In this paper we propose a methodology devoted to exploit high resolution radars for monitoring water bodies in semiarid countries. The proposed approach is based on appropriate registration, calibration and processing of SAR data, producing information ready to use by end-users. The obtained results were used to (i) estimate a relationship between surface and volume of water stored in reservoirs and (ii) validate a hydrological model that simulates the time evolution of water availability.

In this paper we present the Multitemporal Adaptive Processing (MAP3) framework for the definition of a new family of multitemporal, user-oriented products whose information level lays between those of the already available Level 1 and Level 2 products. This framework is organized in three blocks of activities dealing with pre-processing, adaptive processing and representation. Experiments performed on semiarid and temperate datasets testify the reliability of the proposed framework and its independence from the sensor and the scenario.

Reliable segmentation of SAR images requires some forms of user supervision: we resort here to the interactive version of the Tree-Structured Markov Random Field (TS-MRF) segmentation suite. The TS-MRF model, and the associated segmentation tool, provide a flexible and spatially adaptive description of the data. In the interactive version, the user can drive the process based on the inspection of the current result, deciding step-by-step which direction to take, and switching from one segmentation modality to another. Experiments with the segmentation and classification of multitemporal SAR images prove the potential of the interactive approach and of the TS-MRF tool.

We address the problem of SAR despeckling by resorting to nonlocal filtering guided by an optical image. In fact, given the increasing availability of remote-sensing optical images, it makes perfect sense trying to use them to improve the performance of despeckling. Our technique exploits the optical image to reliably estimate the statistical similarity among pixels, which is used to evaluate the weights of nonlocal filtering. Optical data are not used to estimate SAR values, but only to guide the overall process. In addition, they are discarded altogether in regions where SAR and optical images present different local geometries, identified by a preliminary classification step, avoiding thus any additional distortion. Experimental results show the proposed approach to provide images of better quality than state-of-the-art conventional filters.

High resolution SAR data can be a powerful support mainly in areas where the acquisition of in situ information is hampered by physical or economic obstacles. Purpose of this paper is to present an approach to exploit high resolution SAR data for monitoring the temporal evolution of reservoir characteristics in semi-arid regions. Classical and innovative techniques are tailored on the specific climatic conditions of these regions, characterized by the alternation of a three months wet and a nine months dry seasons. Results from a case study developed in Burkina Faso show that the combined use of amplitude and phase information allows the estimation of the eroded areas and a meaningful monitoring of the reservoirs sedimentation.