Real-time fire monitoring based on continuous acquisitions of EO images and geospatial data
We have developed a wildfire monitoring service that exploits satellite images and linked geospatial data to detect fire hotspots and monitor the evolution of fire fronts in real-time. The service makes heavy use of scientific database technologies (array databases, SciQL, data vaults) and linked data technologies (ontologies, linked geospatial data, stSPARQL) and is implemented on top of the state-of-the-art DBMS MonetDB and Strabon. The architecture of the service is depicted in the figure bellow and its workflow is described as follows.
A ground based receiving antenna receives every 5 and 15 minutes satellite images originating from the SEVIRI (Spinning Enhanced Visible and Infrared Imager) sensor on top of the METEOSAT Second Generation satellites MSG-1 and MSG-2. The data vault is responsible for the ingestion policy and enables the efficient access to large archives of image data and metadata in a fully transparent way, without worrying about their format, size and location.
Using the data vault the satellite images are loaded in MonetDB and a hotspot detection processing chain is applied to them. The processing chain is implemented in a straightforward manner using the scientific database language SciQL.
The output of this processing chain is a set of rectangles (pixes of the images) classified as fire, potential fire or no fire. The fire pixels derived by the above processing chain have dimensions equal to the sensor’s spatial resolution, in this case nearly 4x4 km. MSG/SEVIRI is a low resolution observational system, compared to other sensors. However, the unique advantage of MSG/SEVIRI is its geostationary orbit, which allows for a very high observational frequency (5-15 minutes) over the same area of interest. Due to the low spatial pixel resolution of the MSG/SEVIRI instrument and errors in image geo-referencing some false alarms and omission errors exist in the hotspot detection processing chain. For example, some hotspots wrongly appear to be over inconsistent underlying land use/land cover classes. Such kind of errors should be corrected in a clear and systematic way, to ensure the reliability and transferability of the service to other geographic areas.
The generated products produced by this processing chain are annotated semantically, according to a geospatial ontology, and transformed into RDF to be combined with linked geospatial data (e.g., CORINE Land Cover, Greek Administrative Geography, GeoNames) and to offer to users functionalities that go beyond the ones currently available to them.
For example, an additional processing step refines the RDF representation of the generated products taking into account linked geospatial data available on the Web and it improves the accuracy of standard products by correcting the false alarms and omission errors described earlier.
Finally the front-end interface enables controlling the back-end functionality with user-friendly tools, and disseminating the products to the end-user community. The fire monitoring service is publically available . The service is password protected so if you want to access it please contact the National Observatory of Athens which is the TELEIOS partner responsible for administrating the service. The service has been tested extensively by users during the user community workshops organized by TELEIOS. It was very well received and the feedback collected was very encouraging. Also, a demo of the service has been submitted to the Semantic Web Challenge 2012 which took place in Boston in the context of the International Semantic Web Conference and it was awarded 3rd place in the challenge (http://challenge.semanticweb.org/2012/winners.html).
See the deliverables of WP7 and the demo paper Real Time Fire Monitoring Using Semantic Web and Linked Data Technologies for more details.
Recent News: The real-time fire monitoring and burnt scar mapping applications built in TELEIOS are the basis for the FIREHUB service which was awarded the Best Service award in the Copernicus Master's competition of 2014.