Smart Information Instead of „Big Data“

We are surrounded by a multitude of sensors that permanently measure the environment – and produce a real flood of data. For example, modern geo-information is based on satellite measurements, aerial photographs, microscopic and 3D representations that provide data on a wide variety of aspects.

Given a specific question – for example, if one wants to identify all inundated areas after a flood – it is often very time-consuming to “extract” the desired answer from the surplus of data.

In the MENELAOS_NT research project, scientists from the Center for Sensor Systems (ZESS) at the University of Siegen, together with international partners, aim to develop a novel sampling paradigm enabling the generation of smart information directly from sensor data.

The project is funded by the European Commission within the framework of the EU program for research and innovation “Horizon 2020” with around 3.7 million euros.

It is one of the largest research projects at the University of Siegen. MENELAOS_NT is also designed as a training network for a total of 15 doctoral positions, three of them in Siegen.

“Due to the enormous development in the field of sensor technology, we now have not only a huge amount, but also very detailed data at our disposal – more than we can even analyze,” says Prof. Dr. Otmar Loffeld, Chairman of ZESS and Coordinator of MENELAOS_NT. Loffeld considers the “big data” approach impractical, in which all data is first captured and collected in order to see what information can be obtained from it:

“When you go to the supermarket, you don't first buy all the shelves empty and then decide which products you need for cooking. You first make a list and then select only the food you need. You can imagine our approach at MENELAOS_NT in a similar way.”

For the scientists, it is a matter of using new technologies to specifically capture signals and images in such a way that only interesting information is retained, using very low volumes of data. “A promising approach from mathematics is, in this context, 'compressed sensing'.

This makes it possible to capture signals or other information sources in compressed form – this means that from the outset only a few sensor data with a high information content are recorded,” explains Dr. Miguel Heredia Conde, Group Leader at ZESS and Project Manager of MENELAOS_NT.

The principle of 'compressed sensing' can be illustrated using the example of modern digital cameras: they have an extremely large number of pixels and produce photos in such high resolution that they have to be compressed in the chip afterwards so that the files do not become too large.

The large volume of data recorded by the sensor therefore seems to be not so useful – it would be simpler and more resource-saving to take photos in compressed form. ‘Compressed Sensing' enables such compression with the aid of mathematical calculations. Prof. Loffeld explains: “Think, for example, of the universe as a three-dimensional space.

There are as many positions as you want – but there are only stars at certain positions. With compressed sensing, not all conceivable positions are measured, but only few combinations that are randomly selected. How many measurements are required can be calculated mathematically and is also dependent on how close these relevant positions – in our example, the stars – are to each other.”

Within the framework of MENELAOS_NT, the scientists want to develop a solid method to apply the mathematical principle of “compressed sensing” at all levels of sensor technology, signal processing and information retrieval. The aim is to evaluate and combine the measured values of different sensor systems (optical and non-optical sensors, systems for close and remote sensing) as cleverly as possible in order to obtain focused information with regard to a specific problem.

“We would then have the opportunity to observe fundamental processes of our environment more closely and to understand them better. This would be of great benefit with regard to many current challenges – from climate change to sustainable agriculture and forestry, to the efficient use of resources and the protection of peace and security in Europe,” hopes Loffeld.

But the new method of gathering information would also promise rapid progress for very concrete applications: for example, self-propelled cars could identify obstacles more quickly and more precisely on the basis of improved, three-dimensional information. Gesture recognition in mobile phones and robotics could also be significantly improved.

Background:
In addition to the University of Siegen, ten well-known partners (universities, research institutes and industrial partners) are involved in the Innovative Training Network (ITN) MENELAOS_NT (“Multimodal Environmental Exploration Systems_Novel Technologies”): the Fraunhofer FHR, the German Aerospace Center, the Siegen-based company pmdtechnologies ag, AMO GmbH from Aachen, the Weizmann Institute of Science (Israel), Research Center for Spatial Information (Romania), Sabanci University (Turkey), Centro Singular de Investigación en Tecnoloxías da Información (Spain), Gamma Remote Sensing AG (Switzerland) and Insitu Engineering (Spain). The University of Siegen is in charge of the project which starts in January 2020 and will run for four years.

This project has received funding from the European Union`s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 860370.

Prof. Dr.-Ing. Otmar Loffeld
e-mail: loffeld@zess.uni-siegen.de
phone: 0271-740 3125

Dr.-Ing. Miguel Heredia Conde
e-mail: heredia@zess.uni-siegen.de
phone: 0271-740 2465