You must have a javascript-enabled browser and javacript and stylesheets must be enabled to use some of the functions on this site.
 
        

 

The Use of Knowledge-Driven Information Mining to Provide New Services for Enhanced Use of MERIS Data

Mihai Datcu(1) , Andrea Colapicchioni(2) , Sergio D'Elia(3) , Klaus Seidel(4) , and Domenico Solimini(5)

(1) DLR - German Aerospace Center, Oberpfaffenhofen, D-82234 Wessling, Germany
(2) Advanced Computer Systems SpA (ACS), Via della Buffalotta, 378, I-00139 Roma, Italy
(3) ESA-ESRIN, Via Galileo Galilei, 00044 Frascati, Italy
(4) ETHZ - Swiss Federal Institute of Technology, Gloriastr. 35, CH-8092 Zurich, Switzerland
(5) University of Rome II Tor Vergata, Via di Tor Vergata, I-00133 Roma, Italy

Abstract

In recent years, our capability to acquire and archive large volumes of data has greatly exceeded our capability to extract information and add value to the data. This has led to combined efforts to develop new concepts and methods to deal with large data sets: query by image content, information mining and knowledge discovery.

In this frame, the huge potential of MERIS data stimulated efforts to enlarge the areas of applications, to automatise the information extraction, to enable the easy access of users directly to the information content. One of the prototyped and demonstrated solutions is the advanced user access to the information content of MERIS data through the KIM / KES system (KIM = Knowledge-driven Image information Mining; KES = Knowledge Enabled Services).

The KIM/KES system combines physical parameter retrieval with spatial data analysis, image classification and information mining, as well as content-based queries with GIS and feature extraction, thus giving a new dimension to the information content and knowledge extracted from newly acquired data and to its aggregation with existing information in the database.

The MERIS level 1 product is processed for primitive feature extraction, as a quasi-complete signal characterisation: spectral signatures, texture, multiscan analysis, region extraction and shape parameters. The extracted primitive features are coded and represented as entries in a special catalogue. This catalogue is accessed by machine learning algorithms, enabling the users to rapidly and interactively gather relevant information from large image archives via intelligent visual interfaces.

Operating the KIM/KES system we demonstrated a broad range of information extraction or discovery, for example detection of algae blooms, sediments, assessment of vegetating cover, observation of forest fires, or generation of cloud maps.

Presently the KIM/KES system is being progressively integrated into the MERIS ground segment for operational use. It is used also to create SSE services (SSE = Service Support Environment), which can be activated by users or chained with other services by service providers.

 

Workshop presentation