The research cluster of Big Earth Data Analytics, tailored to EO data, allows the management and presentation of vast amount of EO data and the discovery of new information that is hidden in the data and promote the value-adding combination with non-EO data streams. The cluster will research and develop technologies related to data mining, machine learning, visual spatio-temporal exploration of big geospatial and temporal data, semantic enrichment of EO data and products, fusing EO and crowd-sourced data generated from smart sensor technology and geoinformatics. In the last years, different data mining technologies have been developed to cope with the different volume, variety, velocity and veracity of space-based data in order to make the EO services and applications development more efficient and to benefit from all the information hidden within the data. The need to move geospatial data analysis, and more specifically EO data processing, into the “cloud” and to store and represent data in formats (e.g., data cubes) has been recognised by many organisations worldwide. Consequently, several organisations and initiatives worldwide have already begun or are preparing for the uptake of EO data into their Big Data infrastructures. The activities of the research cluster are presented in more detail in the following sections.
The Big Earth Data Analytics Department consists of 7 researchers (3 Postdoctoral and 4 PhD students) and is coordinated by Mr Gunter Schreier from DLR, one of the EXCELSIOR project’s advance partners.
Information extraction refers to techniques such as data mining, machine learning and semantic annotation to extract actionable knowledge hidden in EO data. It is often used with general analytical methods for the exploitation of the information contained in Time Series satellite images. The main focus is on the information extraction in the form of “categories of evolution” and elaboration of technologies to classify the evolutions processes of observed scenes.
Visual exploration allows interactive data presentation in order to increase users’ capabilities to understand the information content of large data sets of images and extract meaningful, relevant semantic clusters, together with quantitative measurements presented in a suggestive, visual way. Visual exploration provides a preliminary insight into optical or radar data by revealing its semantic structure and quantitative estimations regarding the structure through simple visual representations. The benefit of the technique is that the end-user can make more informed decisions on the feasibility of the desired image processing. 3D visualization techniques provide capabilities for visual analytics of geospatial and time series data and building Augmented Reality application, by using state of the art technologies, thereby enabling flexible and fast interactive visualisation of big multidimensional spatial data through linked views.
The data fusion process, which combines crowd sourced data with multi-modal EO information, can be used to make a high-resolution value-added map representing the environment at the time at which the observations were made. It requires the analysis of large amounts of data and the use of automatic and semiautomatic tools. Crowd sourcing and data fusion techniques are applied to multisource data for increased accuracy of documentation.
Geoinformatics deal with all information infrastructures related to geospatial data, such as acquisition, analysis, processing, evaluation, and visualisation, to facilitate the interpretation, management and decision-making in basic research, as well as the addressing of complex social and environmental challenges. Geo-informatics can be used to identify property, infrastructure and cultural heritage monuments that are damaged as a result of geo-hazards. It includes the techniques of geoprocessing, geographic information systems, geometry computer modelling, coordinate reference systems and frames, precise position techniques and navigation.