Millions of people carry environmental sensors with them, for example, to measure temperature, humidity, or air pressure. Multisensor systems monitor potentially dangerous changes in the environment. Current visions of the future suggest that we will increasingly be able to interpret, evaluate, and clearly communicate environmental changes in near real time. Ideally, this will lead to a sustainable improvement or safeguarding of quality of life, for example in the areas of health, environmental protection, or disaster prevention.

Motivation and Objectives

The goal of this research project is to provide innovative methods that significantly increase the benefits of rapidly growing spatially referenced data sets for the economy and society. For example, a big data analytics system is being developed to address issues in the field of smart cities, focusing on the analysis of various sensor data related to environmental and health issues. Furthermore, the degree of autonomy of optical monitoring systems for the precision monitoring of large structures—such as wind turbines or bridges—is being increased through image analysis using deep learning systems; these systems are being tested for reliability and optimized for practical applicability. The applicability of the results will be ensured through the provision of prototypes.

Specific scientific goals are being pursued in the following sub-areas:

• A big data analytics system for spatial data with method selection —to make new analytical methods available to users without in-depth data science knowledge: Data mining and machine learning methods for large, heterogeneous datasets with spatiotemporal references will be provided. An approach will be developed that evaluates possible analysis methods and recommends them for execution.

• Lifelong machine learning, online learning —to enable self-learning, unsupervised accumulation of knowledge: For selected application areas, data mining and ML methods are being extended to reuse knowledge from previous analysis tasks for new, thematically related investigations.

• Image analysis with novelty detection – as a core element of future monitoring systems for technical infrastructure: As a prominent special case of lifelong ML, the suitability of deep learning for technical monitoring—e.g., of bridges or wind turbines—will be investigated.

• Visual analytics methods —to understand and optimize analysis results and present them in a way tailored to specific target audiences. Methods from the field of geo-visual analytics are being developed to enable users to understand and professionally interpret the results of data mining and ML procedures, and to further refine investigations in accordance with specified objectives.

Activities and Results

Specific scientific goals are being pursued in the following subfields: An interdisciplinary research team at Mainz University of Applied Sciences is dedicated to these goals and is drawing on so-called big data methods, as they are currently emerging in a world of rapidly growing and increasingly heterogeneous mass data. Working in close collaboration, the Geoinformatics division in the School of Engineering, together with i3mainz(the Institute for Spatial Information and Measurement Technology) and the Big Data Analytics division in the School of Business, is further developing promising methods for monitoring natural and human-induced environmental changes. The project investigates the potential of current data mining and machine learning methods for spatiotemporal problems. By developing a meta-learning system—that is, combining predictions from multiple models with novel visualization methods—the project aims to significantly expand the pool of potential users of complex analyses.

Project Leadership

Prof. Dr.-Ing. Martin Schlüter
Prof. Dr. habil. Gunther Piller (Department of Economics)
Prof. Dr.-Ing. Klaus Böhm
Prof. Dr.-Ing. Jörg Klonowski

People Involved

Nicole Bruhn, M.A.
Alexander Rolwes, M.Sc.
Kira Zschiesche, M.Sc.
Denise Becker, M.Sc.
Thomas Müller, M.Sc. (Department of Economics)
Lisa Mosis (Geoinformatics and Surveying)
Linda Rau, B.Sc. (Geoinformatics and Surveying)

Duration

3 years

Funding Agency

Carl Zeiss Foundation

Funding Program

“Transfer” Funding Line for Universities of Applied Sciences 2018

Funding Focus

Digitalization: Researching the Fundamentals—Utilizing Applications

Project Administrator

Carl Zeiss Foundation – Administrative Office

Contact

martin.schlueter (at) hs-mainz.de