Special Issue "Big Data in Earth Observation: A New Computing Paradigm for Remote Data Analysis"

A special issue of Remote Sensing (ISSN 2072-4292). This special issue belongs to the section "Environmental Remote Sensing".

Deadline for manuscript submissions: 10 August 2021.

Special Issue Editors

Dr. Juan M. Haut
E-Mail Website1 Website2 Website3
Guest Editor
Department of computer technology and communications, Polytechnic School of Cáceres, University of Extremadura, 10003 Cáceres, Spain
Interests: hyperspectral image analysis; machine (deep) learning; neural networks; multisensor data fusion; high performance computing; cloud computing
Special Issues and Collections in MDPI journals
Dr. Mercedes E. Paoletti
E-Mail Website
Guest Editor
Department of computer technology and communications, Polytechnic School of Cáceres, University of Extremadura (avenida de la Universidad s/n, 10003, Cáceres CÁCERES, Spain)
Interests: hyperspectral remote sensing; deep learning; Graphics Processing Units (GPUs); High Performance Computing (HPC) techniques
Special Issues and Collections in MDPI journals
Dr. Zebin Wu
E-Mail Website
Guest Editor
School of computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China
Interests: hyperspectral image processing; remote sensing big data processing; parallel computing; machine learning; cloud computing
Special Issues and Collections in MDPI journals

Special Issue Information

Dear Colleagues,

With the recent advances made in the Earth Observation (EO) field, the use of remote sensing information captured by available sensors (located on aerial and/or satellite platforms) has acquired a very important role in a wide range of human activities such as the management of environment and natural resources (including forests, water, geological and mineralogical resources), prevention of risks and catastrophes, planning of urban and rural spaces, detection of military objectives and intelligence tasks, among others. This has been fostered by the fact that a detailed characterization of the Earth's surface is now possible using the data collected by current remote sensing instruments for EO, which are able to collect data with higher spatial and spectral resolutions, thus allowing for the acquisition of a large variety of remotely sensed images, from panchromatic and RGB data to multispectral and hyperspectral scenes, from LiDAR and radar sensors, to thermal and optical images, and from low to medium, high and very high spatial resolutions.

For instance, the sensors capable of acquiring images with hundreds of spectral bands (called imaging spectrometers) are able to gather large amounts of information for the same area by recording hundreds of measurements in the spectral domain at different wavelengths. This allows "to see what the human eye cannot," making possible the generation of "data cubes," also known as hyperspectral images (HSI) with very large dimensionality. These images permit a very precise characterization of the terrestrial surface. For example, NASA's Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) sensor is able to capture HSI scenes with 224 spectral bands between 0.4 and 2.5 micrometers, and spatial resolution of about 20 meters per pixel. Such wealth of spatial and spectral information (despite imposing important computational requirements) has opened new possibilities in many applications, including the detailed characterization of agricultural and urban areas, or the monitoring and prevention of natural disasters such as forest fires, oil spills and other types of chemical pollution.

This Special Issue on “Big Data in Earth Observation: a new computing paradigm for remote data analysis" is intended to introduce the latest techniques in high performance computing (HPC) to the development and application of new image processing techniques for an adequate and computationally efficient exploitation of remotely sensed scenes from a Big Data point of view, exploring new computationally efficient models for extracting information from huge remote sensing datasets, with particular interest in the development of parallel and distributed techniques based on graphical processing units (GPUs) and grid/cloud computing platforms.

The goal of this Special Issue is to collect the latest and most advanced ideas regarding the new and efficient techniques for extracting information based on the new trends in advanced learning algorithms (including the newest machine and deep learning approaches).

Dr. Juan M. Haut
Ms. Mercedes E. Paoletti
Dr. Zebin Wu
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All papers will be peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Remote Sensing is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • Big data
  • Neural Networks
  • Deep learning
  • Cloud computing
  • GPUs
  • Heterogeneous computing
  • Remote sensing
  • Supercomputing
  • Image processing
  • Machine learning
  • High performance computing

Published Papers (4 papers)

Order results
Result details
Select all
Export citation of selected articles as:

Research

Article
Sentinel-1 Big Data Processing with P-SBAS InSAR in the Geohazards Exploitation Platform: An Experiment on Coastal Land Subsidence and Landslides in Italy
Remote Sens. 2021, 13(5), 885; https://doi.org/10.3390/rs13050885 - 26 Feb 2021
Cited by 2 | Viewed by 1422
Abstract
The growing volume of synthetic aperture radar (SAR) imagery acquired by satellite constellations creates novel opportunities and opens new challenges for interferometric SAR (InSAR) applications to observe Earth’s surface processes and geohazards. In this paper, the Parallel Small BAseline Subset (P-SBAS) advanced InSAR [...] Read more.
The growing volume of synthetic aperture radar (SAR) imagery acquired by satellite constellations creates novel opportunities and opens new challenges for interferometric SAR (InSAR) applications to observe Earth’s surface processes and geohazards. In this paper, the Parallel Small BAseline Subset (P-SBAS) advanced InSAR processing chain running on the Geohazards Exploitation Platform (GEP) is trialed to process two unprecedentedly big stacks of Copernicus Sentinel-1 C-band SAR images acquired in 2014–2020 over a coastal study area in southern Italy, including 296 and 283 scenes in ascending and descending mode, respectively. Each stack was processed in the GEP in less than 3 days, from input SAR data retrieval via repositories, up to generation of the output P-SBAS datasets of coherent targets and their displacement histories. Use-cases of long-term monitoring of land subsidence at the Capo Colonna promontory (up −2.3 cm/year vertical and −1.0 cm/year east–west rate), slow-moving landslides and erosion landforms, and deformation at modern coastal protection infrastructure in the city of Crotone are used to: (i) showcase the type and precision of deformation products outputting from P-SBAS processing of big data, and the derivable key information to support value-adding and geological interpretation; and (ii) discuss potential and challenges of big data processing using cloud/grid infrastructure. Full article
Show Figures

Graphical abstract

Article
A Parallel Unmixing-Based Content Retrieval System for Distributed Hyperspectral Imagery Repository on Cloud Computing Platforms
Remote Sens. 2021, 13(2), 176; https://doi.org/10.3390/rs13020176 - 06 Jan 2021
Cited by 1 | Viewed by 549
Abstract
As the volume of remotely sensed data grows significantly, content-based image retrieval (CBIR) becomes increasingly important, especially for cloud computing platforms that facilitate processing and storing big data in a parallel and distributed way. This paper proposes a novel parallel CBIR system for [...] Read more.
As the volume of remotely sensed data grows significantly, content-based image retrieval (CBIR) becomes increasingly important, especially for cloud computing platforms that facilitate processing and storing big data in a parallel and distributed way. This paper proposes a novel parallel CBIR system for hyperspectral image (HSI) repository on cloud computing platforms under the guide of unmixed spectral information, i.e., endmembers and their associated fractional abundances, to retrieve hyperspectral scenes. However, existing unmixing methods would suffer extremely high computational burden when extracting meta-data from large-scale HSI data. To address this limitation, we implement a distributed and parallel unmixing method that operates on cloud computing platforms in parallel for accelerating the unmixing processing flow. In addition, we implement a global standard distributed HSI repository equipped with a large spectral library in a software-as-a-service mode, providing users with HSI storage, management, and retrieval services through web interfaces. Furthermore, the parallel implementation of unmixing processing is incorporated into the CBIR system to establish the parallel unmixing-based content retrieval system. The performance of our proposed parallel CBIR system was verified in terms of both unmixing efficiency and accuracy. Full article
Show Figures

Graphical abstract

Article
Fine-Tuning Self-Organizing Maps for Sentinel-2 Imagery: Separating Clouds from Bright Surfaces
Remote Sens. 2020, 12(12), 1923; https://doi.org/10.3390/rs12121923 - 14 Jun 2020
Viewed by 1099
Abstract
Removal of cloud interference is a crucial step for the exploitation of the spectral information stored in optical satellite images. Several cloud masking approaches have been developed through time, based on direct interpretation of the spectral and temporal properties of clouds through thresholds. [...] Read more.
Removal of cloud interference is a crucial step for the exploitation of the spectral information stored in optical satellite images. Several cloud masking approaches have been developed through time, based on direct interpretation of the spectral and temporal properties of clouds through thresholds. The problem has also been tackled by machine learning methods with artificial neural networks being among the most recent ones. Detection of bright non-cloud objects is one of the most difficult tasks in cloud masking applications since spectral information alone often proves inadequate for their separation from clouds. Scientific attention has recently been redrawn on self-organizing maps (SOMs) because of their unique ability to preserve topologic relations, added to the advantage of faster training time and more interpretative behavior compared to other types of artificial neural networks. This study evaluated a SOM for cloud masking Sentinel-2 images and proposed a fine-tuning methodology to separate clouds from bright land areas. The fine-tuning process which is based on the output of the non-fine-tuned network, at first directly locates the neurons that correspond to the misclassified pixels. Then, the incorrect labels of the neurons are altered without applying further training. The fine-tuning method follows a general procedure, thus its applicability is broad and not confined only in the field of cloud-masking. The network was trained on the largest publicly available spectral database for Sentinel-2 cloud masking applications and was tested on a truly independent database of Sentinel-2 cloud masks. It was evaluated both qualitatively and quantitatively with the interpretation of its behavior through multiple visualization techniques being a main part of the evaluation. It was shown that the fine-tuned SOM successfully recognized the bright non-cloud areas and outperformed the state-of-the-art algorithms: Sen2Cor and Fmask, as well as the version that was not fine-tuned. Full article
Show Figures

Graphical abstract

Article
Estimating Water pH Using Cloud-Based Landsat Images for a New Classification of the Nhecolândia Lakes (Brazilian Pantanal)
Remote Sens. 2020, 12(7), 1090; https://doi.org/10.3390/rs12071090 - 28 Mar 2020
Cited by 2 | Viewed by 1522
Abstract
The Nhecolândia region, located in the southern portion of the Pantanal wetland area, is a unique lacustrine system where tens of thousands of saline-alkaline and freshwater lakes and ponds coexist in close proximity. These lakes are suspected to be a strong source of [...] Read more.
The Nhecolândia region, located in the southern portion of the Pantanal wetland area, is a unique lacustrine system where tens of thousands of saline-alkaline and freshwater lakes and ponds coexist in close proximity. These lakes are suspected to be a strong source of greenhouse gases (GHGs) to the atmosphere, the water pH being one of the key factors in controlling the biogeochemical functioning and, consequently, production and emission of GHGs in these lakes. Here, we present a new field-validated classification of the Nhecolândia lakes using water pH values estimated based on a cloud-based Landsat (5 TM, 7 ETM+, and 8 OLI) 2002–2017 time-series in the Google Earth Engine platform. Calibrated top-of-atmosphere (TOA) reflectance collections with the Fmask method were used to ensure the usage of only cloud-free pixels, resulting in a dataset of 2081 scenes. The pH values were predicted by applying linear multiple regression and symbolic regression based on genetic programming (GP). The regression model presented an R2 value of 0.81 and pH values ranging from 4.69 to 11.64. A lake mask was used to extract the predicted pH band that was then classified into three lake classes according to their pH values: Freshwater (pH < 8), oligosaline (pH 8–8.9), and saline (≥9). Nearly 12,150 lakes were mapped with those with saline waters accounting for 7.25%. Finally, a trend surface map was created using the ALOS PRISM Digital Surface Model (DSM) to analyze the correlation between landscape features (topography, connection with the regional drainage system, size, and shape of lakes) and types of lakes. The analysis was in consonance with previous studies that pointed out that saline lakes tend to occur in lower positions compared to freshwater lakes. The results open a relevant perspective for the transfer of locally acquired experimental data to the regional balances of the Nhecolândia lakes. Full article
Show Figures

Graphical abstract

Planned Papers

The below list represents only planned manuscripts. Some of these manuscripts have not been received by the Editorial Office yet. Papers submitted to MDPI journals are subject to peer-review.

(1) Combining Distributed Spatial Computation and a Parallel Evolutionary Algorithm for Fast Urban LiDAR Flight Path Planning

Anh Vu Vo1, Debra F. Laefer2, Jonathan Byrne3

1 University College Dublin, Ireland; [email protected]
2 New York Univeristy, USA; [email protected]edu
3 Intel Movidius, Ireland; [email protected]

The use of parallel and distributed computing in the field of Earth bservation is well recognised in the context of post-acquisition data analysis and processing. Namely, coupling a large number of computing processors to work concurrently can accelerate computations involving large amounts of data aggregated from decades of sensing activities. In this paper, a different perspective is taken. We investigate the use of parallel and distributed computing for planning data acquisition. Specifically, a computing framework is introduced to facilitate LiDAR flight path planning for dense urban environments.

As the majority of standard practices in airborne LiDAR data acquisition were developed for topographic mapping, mapping dense urban environments requires re-evaluation. For instance, flight line directions must be planned to alleviate occlusions caused by tall features and to maximise data capture on building facades. Such problems do not arise typically in topographic mapping, but they are crucial for comprehensive mapping of dense urban environments.

The flight path planning in this research is formulated as a geometric problem optimised with an evolutionary algorithm, which is capable of optimising open-ended problems that have many possible solutions. The evolutionary optimisation operates by iteratively evaluating a set of candidate flight paths and combining the best performing candidates, which are in part randomly mutated to expand the search space. Evaluating the performance of a large number of flight path candidates repeatedly is a time consuming process. The speed of the process decides the feasibility of the optimisation strategy.

In this research, two levels of parallelisation are combined to curb the required computational time of the LiDAR flight path optimisation. At the first level, multiple flight path candidates of the same generation are evaluated simultaneously. This level of parallelisation is straightforward since evolutionary algorithms are inherently parallelisable. At the second level, the evaluation of each flight path candidate is conducted using a novel distributed algorithm. The algorithm, which is based on the Map Reduce programming model, makes use of multiple computing cores of a shared-nothing cluster to reduce the runtime of each evaluation. The second level of parallelisation is crucial to the reduction of the overall runtime given that the evaluation of each flight scenario requires a complex, intensive computation that involves a large amount of spatial data.

The flight path optimisation strategy is demonstrated through the planning of an actual extremely highresolution LiDAR scanning over an area of 1km2 of the Sunset Park area in Brooklyn, New York. In that project, each flight path evaluation took under 3 minutes, bringing the total time for achieving a converged result to approximately 30 hours. To confirm the validity of the optimisation, the best- and worst-performing flight path candidates were flown in May 2019. The data acquired from the flight provide robust evidence to justify the optimisation strategy and provide important insight into the LiDAR flight planning process for dense urban environments.

 

(2) PyCircularStats: A Python-based tool for circular statistics and graphical analysis

Aurora Cuartero Sáez, Pablo García Rodríguez

Abstract: Circular data, as part of directional data, is used in a wide range of fields, such as Geology, Biology, Meteorology, and Geomatics. It differs from traditional linear data because it is closed and has no beginning or end along the actual line, i.e. circular data occurs around a circle, normally measured in degrees. Analyzing directional data, in particular circular data, requires methods that are available in libraries with a well-known prestige as Python including its SciPy, NumPy or SciKit-Learn libs. However, these libraries have a specific area of expertise and do not combine information in a useful way for two-dimensional data analysis. In this paper, an open-source library has been implemented to be executed by the Python interpreter, called PyCircularStats. To demonstrate the potential of PyCircularStats and to show some of its features, the integration of the proposed circular statistics is described with the objective of analyzing vector data. As an example, a particular case is shown to analyze the positional accuracy of satellite image of LandSat-8 in Caceres, Spain, with this graphical circular statistic. Additionally, this paper performs a comparison between PyCircularStats and VecStatGraphs2D, another vector analysis using graphical and analytical methods developed in R, and also the improvements and advantages developed in a new PycircularStats tool based on Python have also been presented.

Back to TopTop