Tools and Platforms

The upcoming suite of Sentinel satellites in combination with their free and open access data policy will open new perspectives for establishing a spatially and temporally detailed monitoring of the Earth’s surface. The Sentinel fleet will provide a so-far unique coverage with Earth observation (EO) data and new possibilities with respect to the implementation of innovative methodologies, techniques and geo-information products and services. However, the capability to effectively and efficiently access, process, analyze and distribute the mass data streams from the Sentinels and high-level information products derived from them poses a key challenge. This is also true with respect to the necessity of flexibly adapting the processing and analysis procedures to new or changing user requirements and technical developments. Hence, the implementation of operational, modular and highly automated processing chains, embedded in powerful hard- and software environments and linked with effective distribution functionalities, is of central importance.

This contribution introduces the TEP Urban project that aims at the utilization of modern information technology functionalities and services to bridge the gap between the technology-driven EO sector and the information needs of environmental science, planning, and policy. Key components of such a system are currently developed in the TEP Urban project. This includes the implementation of an open, web-based platform employing distributed high-level computing infrastructures (Platform as a Service – PaaS) as well as providing key functionalities for i) high-performance access to thematic data (Information as a Service – InaaS), ii) modular and generic state-of-the art pre-processing, analysis, and visualization (Software as a Service – SaaS), iii) customized development and dissemination of algorithms, products and services, and iv) networking and communication. These services and functionalities are supposed to enable any interested user to easily exploit and generate thematic information on the status and development of the environment based on EO data and technologies.

The TEP Urban platform is supposed to initiate a step change in the use of EO data by providing an open and participatory platform based on modern ICT technologies and services that enables any interested user to easily exploit and generate thematic information on the status and development of the built environment.

The “Citizen Scientists” (CS) concept is one of the emerging fields recognized as a key element to increase EO capabilities. This technology and culture can be used to enable communities of citizens to provide essential information for deeper analysis of remotely sensed information [1]. However, in such a context, metrics and processes for data quality assessments are issues of crucial importance to document the usefulness of the results that can be produced.

In this work a CS project addressing accurate classification of Very High Resolution (VHR) optical imagery, with less than 50 cm spatial resolution, taken over the Rome urban area, Italy, is presented. Besides the data, the “scientists” (students of the University of Rome “Tor Vergata”) have been provided a neural network based toolbox for image processing plus additional scripts to test the accuracy of the obtained results. The area extension includes the main historical city and its surroundings covering more than 70 km².The considered land cover classes are: buildings, asphalted areas, natural, water, bare soil, vegetation. Each tile, with dimension 2000x1000, has been assigned to the “scientist” who mainly has to perform two tasks: image classification and accuracy evaluation. The automatic image classification is performed using the Neumapper Toolbox which is a free distributed package enabling image classification via neural networks [2]. Neural networks have greatcapabilities as a pattern recognition method for multi-source remotely sensed data because of the parallel nature of the processing. In particular, it has been shown that multilayer perceptrons (MLP) maybe an efficient alternative to conventional statistical approaches for automating image classification [3],[4]. With Neumapper the CS can easily manage the whole processing in a unique user friendly software environment. Moreover, the CS is invited to use a few scripts, developed in python language, implementing the following tasks: generation of a given number of validation points randomly distributed over the image, localization of the generated points in Google Earth, realization of a confusion matrix based on the generated (manually labeled) points and the classified image.

An assessment of the global result has been carried out by means of an external validation and it confirmed an overall accuracy around 95%. Even if an on-line assistance service has been also set-up to support the participants in their operations, no particular difficulties have been registered until now. The idea is indeed to extend the project to other urban areas to be mapped so as to yield very precise land cover atlas.

[1] Goochild, M.F., “Citizen as sensors: the world of volunteered geography,” GeoJournal, 69:211-221, 2007

[2] Del Frate, F., I. Fabrini, M. Penalver, M. Iapaolo, “NEUMAPPER: a Neural Networks Software for Image Classification,” Proc. of the ESA-EUSC-JRC 2011 Image Information Mining conference, ISPRA (VA), Italy, 30-31 March, 2011

[3] Pacifici, F., F. Del Frate, W. J. Emery, P. Gamba, J. Chanussot, “Urban mapping using coarse SAR and optical data: outcome of the 2007 GRS-S data fusion contest,” IEEE Geoscience and Remote Sensing Letters, vol 5, n. 3, pp. 331-335, July 2008

[4] Del Frate, F., F. Pacifici, G. Schiavon, C. Solimini, “Use of neural networks for automatic classification from high resolution imagery,” IEEE Transactions on Geoscience and Remote Sensing, vol. 45, n. 4, pp. 800-809, April 2007.

When we look at current developments in Earth Observation and satellite images, we are facing both Big Data issues as well as a high interest in detailed urban mapping. As a consequence, new concepts for exploration and information retrieval are urgently needed. To this end, we propose to explore urban and similar image data via an image information mining approach in which the main steps are feature extraction, classification, semantic annotation, and interactive query processing. In addition, we deal with the integration of multiple sources of information such as synthetic aperture radar (SAR) images and their metadata, semantic descriptors of the image content, as well as other publicly available geospatial data sources expressed as linked open data for posing complex queries in order to support geospatial data analytics.

A recently published approach lays the foundations for the development of powerful analysis tools that focus on urban image interpretation using ontologies and linked open data [1, 2]. We introduced a system architecture where a common satellite image product is transformed from its initial format into to actionable intelligence information, which includes image descriptors, metadata, image tiles, and semantic labels resulting in an EO data model. This opens the way toward an automated identification and classification of urban areas, their infrastructure (e.g., airports), geographic objects (e.g., rivers), and industrial installations. Applications that may result from this work can be a semantic catalogue of urban images to be used in crisis situations or after a disaster.

We also created a SAR image ontology based on our EO data model and a three-level taxonomy classification scheme of the image content. Our approach links, for instance, high-resolution TerraSAR-X images with information from CORINE Land Cover (CLC), Urban Atlas (UA), GeoNames, and OpenStreetMap (OSM), which are represented in the standard triple model of the resource description frameworks (RDFs).

In order to allow interactive querying of images, the stRDF model and the stSPARQL query language have been implemented in the Strabon system, which is freely available as open source software. Strabon extends the well-known open source Sesame 2.6.3 RDF store and uses PostGIS as its spatially enabled backend DBMS.

References:

[1] D. Espinoza-Molina, C. Nikolaou, C.O. Dumitru, K. Bereta, M.  Koubarakis, G. Schwarz, and M. Datcu: Very-High-Resolution SAR Images and Linked Open Data Analytics Based on Ontologies, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 8, pp. 1696-1708, 2015.

[2] C.O. Dumitru, S. Cui, G. Schwarz, and M. Datcu: Information Content of Very-High-Resolution SAR Images: Semantics, Geospatial Context, and Ontologies, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 8, pp. 1635-1650, 2015.

Urban areas, especially in developing countries are the planning hot spots of the future and therefore standardized, accurate and up-to-date geospatial data are needed. Space-based satellite systems can provide very useful information for public authorities and private industry to accurately monitor urban development, to analyze different urban structures and features and to efficiently plan urban infrastructure projects.

GAF AG developed and generates, in close co-operation with the Remote Sensing Technology Institute (IMF) of the German Aerospace Center (DLR) and the data providers Airbus DS and DigitalGlobe, some standard products that can be used for different urban applications.

The Tri-Stereo Digital Surface Model (DSM) as the high-end standard product within GAF’s suite of elevation models is generated from very high resolution (VHR) satellite images with triple stereo coverage (sensor independent). The 0.5 m resolution DSM is produced highly standardized and fully quality controlled and can be generated worldwide. The DSM can be used e.g. for propagation analysis, viewshed calculations and 3D flights to get realistic impressions of the urban structure. The near-nadir ortho image, which is also part of the product, can be used e.g. for building footprint digitalization.

As the generation of 3D building models is certainly possible with this stereo data, but as it is also very time-consuming and expensive, urban block models may be the preferred solution. These models structure built-up areas in cities according to different criteria like building density, -type and –function. Furthermore, height values coming from DSM and DTM files are also derived and are part of the product. Most of these processing steps are carried out highly automated. These urban block models are either generated based on the Tri-Stereo product or by using Euro-Maps 3D elevation models.

Euro-Maps 3D is a 5m resolution digital surface model generated from IRS-P5 Cartosat-1 2.5 m in- flight stereo satellite data. This DSM is mainly used for regional and trans-national issues, but as buildings and building blocks are very sharply visible, it can be also used for urban applications. This product has also been developed in close co-operation with the Remote Sensing Technology Institute (IMF) of the German Aerospace Center (DLR).