Big Spatial Data Analytics
What we do
We live in a time where vast amounts of spatial data are generated by technical sensors, social media users, and volunteers via crowdsourcing. Processing such large data poses complex challenges due to their mere volume and semantic complexity. We help by rendering these datasets usable for your application, while always taking into account the spatial context. Our long experience gathered in numerous research projects enables us to serve as an interface between technology and its application. Based on your specific needs, we develop processes and tools for assessing the quality and enriching heterogeneous Web 2.0 data by applying innovative methods from spatial data mining and deep learning.
OpenStreetMap data are continuously amended, updated, and corrected. All changes are saved to ensure full traceability. The ohsome analysis platform enables easy access to the full history of OpenStreetMap – worldwide and precise to the second. This way, all (historical) OpenStreetMap elements ever recorded can be reconstructed and analyzed. One important objective is the improved assessment of the quality and usability of OpenStreetMap for your individual application.
Big Data Technology
We deploy big data technology and cluster computing to enable parallel data processing on a scalable server cluster.
Integration through API
Our programming interfaces can be integrated into different systems. This feature enables users to implement their own customized analyses based on the OpenStreetMap history.
Intrinsic Quality Assessment
Based on the ohsome platform, we provide intrinsic quality indicators to effectively support the data quality assessment for various applications.
Deep VGI (Deep Learning with Volunteered Geographic Information) connects user-generated geodata with machine-based learning. Learning algorithms are already successfully used in the field of geoinformation; however, the required training data are often scarce, particularly for rural areas and developing countries. This gap is filled by volunteered geographic information, which helps us optimize the machine-based identification of buildings on satellite images.
Wealth of Information
We use OpenStreetMap data and MapSwipe data due to their abundance and comprehensiveness, making them an ideal foundation for improving the precision of machine-based learning algorithms. We extract relevant information and gain a better understanding of spatial structures and processes.
Crowd und Machine
Analysis tasks involving the automated gathering of information support users and free human analysis skills for tasks that can’t be automated. The consequent savings of time and resources is particularly relevant in time-sensitive situations, such as the mapping efforts for destroyed infrastructure in the aftermath of disaster events.
Improve Data Quality
Machine-based learning is another option for analyzing geodata quality. The automatically generated results help us better understand the user-generated geodata, and to quickly and reliably identify areas with “good” and “poor” geodata.
OSM Landuse Landcover is a WebGIS application to explore the OpenStreetMap database specifically in terms of landuse and landcover information. This is based on our earlier work on testing the suitability of OpenStreetMap for deriving landuse and landcover information.
Visualizing metadata (information about the data) not usually displayed in OpenStreetMap maps offers important intelligence in regards to the data quality. The web application OSMatrix provides such information rendered in a hexagonal raster, for instance detailing timeliness and completeness of OpenStreetMap data. The visual analysis of the resulting maps provides a first assessment of the data quality.
The Nepal Dashboard provides a tailor-made analysis platform based on the ohsome platform. In consultation with one of our partners, Kathmandu Living Labs, we focus on analyzing the mapping activities in Nepal. In the future, the dashboard is designed to visualize information about user-specific activities. This way, we want to enable the early detection of potential mapping problems, connect the individual mappers, and evaluate the effect of training measures. These objectives were heavily influenced by the experience with the thousands of mappers who contributed to the OpenStreetMap efforts in the aftermath of the earthquakes of April/Mai 2015.
To get an impression of the impact that the work of thousands of volunteers had on OSM data in the Nepal region in 2015, a spatio-temporal distribution of contibutor activity has been computed. This visualization is based on the ohsome platform.
HistOSM is a WebGIS application to visually explore historic objects stored in the OpenStreetMap database. Your visual exploration process is supported by dynamically created statistics showing the top most used object categories in the current map view extent.