The present contribution focuses on applying Denoising Diffusion Probabilistic Models to Remote Sensing image classification, generation and super-resolution. Diffusion models are enhanced, including an attention block embedded in a UNet architecture is used to generate images to complement the EuroSAT data set. Furthermore, models with the same architecture super-resolve sentinel-2 optical images. The results indicate that diffusion models with attention can provide a promising methodological path for applications such as estimating the NDVI images most probably associated with given electro-optical and SAR acquisitions thereby overcoming limitations in observability due to cloud cover or solar illumination.

The creation of image classification and segmentation Machine Learning products requires the annotation of different classes by experts as well as the management of large volumes of data. This paper introduces a new methodology to optimize the computational execution and expert-user contribution by introducing a pixel quality indicator and, therefore reducing the number of annotated data used for model training, based on the geometric information of each pixel. The developed pixel-level quality indicator shows beneficial results, as a result of improving semantic segmentation and classification tasks’ performance, validated through the IRIS1platform.

The Normalized Difference Vegetation Index (NDVI), an indicator of vegetation health, is derived from the visible and near-infrared light reflected by vegetation, which can be measured with multi-spectral sensors. However, clouds can often obstruct land areas, making it challenging to obtain NDVI maps of the surface. This study explores the possibility of estimating NDVI from synthetic aperture radar (SAR) images using a Conditional Generative Adversarial Network (cGAN). Results encourage using a cGAN to address the task and provide valuable insights for future improvements.

The correct functioning of photovoltaic (PV) cells is critical to ensuring the optimal performance of a solar plant. Anomaly detection techniques for PV cells can result in significant cost savings in operation and maintenance (O&M). Recent research has focused on deep learning techniques for automatically detecting anomalies in electroluminescence images. Automated anomaly annotations can improve current O&M methodologies and help develop decision-making systems to extend the life cycle of the PV cells and predict failures. This article addresses the lack of anomaly segmentation annotations in the literature by proposing a combination of state-of-the-art data-driven techniques to create a golden standard benchmark. The proposed method stands out for: 1) its adaptability to new PV cell types; 2) cost-efficient fine-tuning; and 3) leverage public datasets to generate advanced annotations. The methodology has been validated in the annotation of a widely used dataset, obtaining a reduction of the annotation cost by 60%.

We present a strategy to create an online phenology prediction service for non-technical users based on machine learning. The presented methodology is explained into two interrelated parts. The first part describes the core of the service, the automated machine learning procedure that enables the prediction modeling of the phenology phases evolution. The second part summarizes the architecture of the pipeline of the service (from data acquisition to interactive interface) considered for the proof of concept. Finally, the olive phenology prediction service developed in the DEMETER H2020 European project is presented as a real usage example.

Using on-demand processing pipelines to generate virtual geospatial products is beneficial to optimizing resource management and decreasing processing requirements and data storage space. Additionally, pre-processed products improve data quality for data-driven analytical algorithms, such as machine learning or deep learning models. This paper proposes a method to integrate virtual products based on integrating open-source processing pipelines. In order to validate and evaluate the functioning of this approach, we have integrated it into a geo-imagery management framework based on Open Data Cube (ODC). To validate the methodology, we have performed three experiments developing on-demand processing pipelines using multi-sensor remote sensing data, for instance, Sentinel-1 and Sentinel-2. These pipelines are integrated using open-source processing frameworks.

Warehouse Management Systems have been evolving and improving thanks to new Data Intelligence techniques. However, many current optimizations have been applied to specific cases or are in great need of manual interaction. Here is where Reinforcement Learning techniques come into play, providing automatization and adaptability to current optimization policies. In this paper, we present Storehouse, a customizable environment that generalizes the definition of warehouse simulations for Reinforcement Learning. We also validate this environment against state-of-the-art reinforcement learning algorithms and compare these results to human and random policies.

We propose a methodology to manage and process remote sensing and geo-imagery data for non-expert users. The proposed system provides automated data ingestion and manipulation capability for analytical data-driven purposes. In this paper, we describe the technological basis of the proposed method in addition to describing the tool architecture, the inherent data flow, and its operation in a specific use case to provide statistical summaries of Sentinel-2 regions of interest corresponding to the cultivation polygonal areas located in the Basque Country (ES).

Industrial process control systems try to keep an output variable within a given tolerance around a target value. PID control systems have been widely used in industry to control input variables in order to reach this goal. However, this kind of Transfer Function based approach cannot be extended to complex processes where input data might be non-numeric, high dimensional, sparse, etc. In such cases, there is still a need for determining the subspace of input data that produces an output within a given range. This paper presents a non-stochastic approximation to determine input values for a mathematical function or trained regression model given an output range. The proposed method creates a synthetic training data set of input combinations with a class label that indicates whether the output is within the given target range or not. Then, a decision tree classifier is used to determine the subspace of input data of interest. This method is more general than a traditional controller as the target range for the output does not have to be centered around a reference value and it can be applied given a regression model of the output variable, which may have categorical variables as inputs and may be high dimensional, sparse... The proposed method is validated with a proof of concept on a real use case where the quality of a lamination factory is established to identify the suitable subspace of production variable values.

Most EO image analysis algorithms ignore existing maps for solving ill-posed scene understanding problems and developing inversion procedures, yet new EO data often represent aspects of known objects and events. To solve part of this problem, we introduce a prototype designed to integrate, process and distribute existing heterogeneous local public authority maps with the outputs of automatic reconstruction/interpretation engines using mathematical tools for geoprocessing and operating on satellite images.

The Petabyte-scale data volumes in Earth Observation (EO) archives are not efficiently manageable with serial processes running on large isolated servers. Distributed storage and processing based on `big data' cloud computing frameworks needs to be considered as a part of the solution.

Image mining consists of the procedures that allow to access, search and explore very large databases of data. Institutions like spatial agencies have to manage huge archives of Earth Observation (EO) images and need solutions to make data available to users from both the algorithmic and the infrastructural point of views. On the other side, users would need to explore the variety of images not just based on metadata, like time of acquisition or sensor parameters, but also by getting knowledge of their content. In this contribution, we investigate methodologies for content-based EO image retrieval via example-based queries. In particular, we present a procedure for the indexing of large-scale unstructured archives, built on top of a cluster analytics framework, Apache Spark. The procedure is based on a hierarchical and scalable implementation of a space partitioning algorithm and allows O(log n) response query times. Scalability analyses are conducted on polarimetric data from NASA/JPL archives, by using virtualized computing resources distributed over the Internet. In particular, the effects of the cluster size and of the hardware scale-up are demonstrated. The results also reveal the applicative potential of using on-demand cloud-based resources.

As per the 2013 EOSDIS annual metrics report1, Petabyte-scale Earth Observation (EO) raster data archive volumes are growing at rates of about ten Gigabytes per day while around 95% of their contents have never been accessed by a human observer [1]. Metadata-based search clearly needs to be complemented by semi-automatic raster catalog content mining.

Big Data Era in Sky and Earth Observation (BIG-SKY-EARTH, http://www.bigskyearth.eu) is a recently started COST Action that aims at setting the ground for a long-term networking between astronomy and remote sensing research communities in the area of Big Data utilization. COST Action is a networking funding scheme that allows researchers to jointly develop their ideas and initiatives in a given scientific field. Even though COST does not fund research itself, COST Actions are active through a range of research supporting networking tools, such as meetings, workshops, conferences, training schools, short-term scientific missions (short travels of researchers to foster their collaborations) and dissemination activities. COST Actions are open to researchers from universities, public and private research institutions, as well as to NGOs, industry and SMEs.

Context categorization is a fundamental pre-requisite for multi-domain multimedia content analysis applications. Most feature extraction methods require prior knowledge to decide if they are suitable for a specific domain and to optimize their input parameters. In this paper, we introduce a new color image context categorization method (DITEC) based on the trace transform. The problem of dimensionality reduction of the obtained trace transform signal is addressed through statistical descriptors of its frequency representation that keep the underlying information. We also analyze the distortions produced by the parameters that determine the sampling of the discrete trace transform. Moreover, Feature Subset Selection (FSS) is applied to both, improve the classification performance and compact the final length of the descriptor that will be provided to the classifier. These extracted features offer a highly discriminant behavior for content categorization without prior knowledge requirements. The method has been experimentally validated through two different datasets.

This paper presents an implementation of query by example in Earth Observation image archive using data compression-based approach. Data compression approach allows to exploit the compression properties of the objects and to estimate the shared information between them, this concept is extended to image retrieval for finding similar objects in the image archive. Our implementation is based on LZW algorithm for compressing the image content and extracting features of the images. The fast compression distance (FCD) is defined as a similarity metric in order to retrieve the most similar images. This tool is satisfactory implemented and tested using optical and SAR images.

Image understanding and retrieval for metric SAR data needs to be tackled by specific algorithms in order to exploit the specific metric scale scene characteristics and the typical sensor phenomenology. A composite approach is developed to this end. Extraction, characterization and simplification/grouping algorithms are employed together with super-resolution for isolated bright scatterers, scatterer alignments and uniform backscatter and texture areas to populate a geospatial database. The generated descriptors can be navigated via different data viewers for scene understanding and data retrieval applications. Examples on real metric SAR data are given.

To investigate the limits and merits of information extraction from a single high-resolution synthetic aperture radar (SAR) backscatter image, we introduce a model-based algorithm for the automatic reconstruction of building areas from single-observation meter-resolution SAR intensity data. The reconstruction is based on the maximum a posteriori estimation by Monte Carlo methods of an optimal scene that is modeled as a set of mutually interacting Poisson-distributed marked points describing parametric building objects. Each of the objects can be hierarchically decomposed into a collection of radiometrically and geometrically specified object facets that in turn get mapped into data features by ground-to-range projection and inverse Gaussian statistics. The detection of the facets is based on a likelihood ratio. Results are presented for airborne data with resolutions in the range of 0.5-2 m on urban scenes covering agglomerations of buildings. To achieve robust results for building reconstruction, the integration with data from other sources is needed.

The exploitation of metric resolution SAR data for the reconstruction of the structure of the observed scenes poses specific problems related both to the complexity of acquired scene details and to the peculiarities of the SAR acquisition system. On the one hand, much more complexity is transferred through the system from the scene into the data: new kinds of complex man-made scene objects are acquired. Layover and shadowing and responses from single scatterers tend to dominate the data. On the other hand, multiple signal reflections, sidelobe effects, radiometric pollution and many other effects related to the increased resolution of the system have to be taken into account. We show how, by properly modelling in stochastic terms the peculiarities of both the acquisition system and of the scene and by composing them in a Bayesian framework, new methods are developed that allow the reconstruction of the imaged structures from SAR data. Particular interest is devoted to the application of the developed algorithms in urban environments on data resolutions ranging from a few metres to fifty centimetres.

In this paper, we demonstrate the concepts of a prototype of a knowledge-driven content-based information mining system produced to manage and explore large volumes of remote sensing image data. The system consists of a computationally intensive offline part and an online interface. The offline part aims at the extraction of primitive image features, their compression, and data reduction, the generation of a completely unsupervised image content-index, and the ingestion of the catalogue entry in the database management system. Then, the user's interests-semantic interpretations of the image content-are linked with Bayesian networks to the content-index. Since this calculation is only based on a few training samples, the link can be computed online, and the complete image archive can be searched for images that contain the defined cover type. Practical applications exemplified with different remote sensing datasets show the potential of the system.

We present a framework for scene understanding from interferometric synthetic aperture radar data that is based on Bayesian machine learning and information extraction and fusion. A generic description of the data in terms of multiple models is automatically generated from the original signals. The obtained feature space is then mapped to user semantics representing urban scene elements in a supervised step. The procedure is applicable at multiple scales. We give examples of urban area classification and building recognition of Shuttle Radar Topography Mission data and of building reconstruction from submetric resolution Intermap data.