Image Information Mining in Time Series

STATUS: Finished

The Image Information Mining in Time Series [1] (IIM-TS) is a 3 year ESA project started in February 2007. The main objective of the project is the extension of KEO (Knowledge-centred Earth Observation) earth observation image information mining system [2] for processing of time series of images. The project is carried out by a consortium led by Italian company Advanced Computer Systems, S.p.A. (prime contractor) which won the bid for project subject to Invitation to Tender AO/1-5119/06/I-OL. Beside the Czech company Iguassu Software Systems, a.s. (ISS), the members of the consortium are: CNES (France), DLR (Germany), ENST (France), Sarmap s.a. (Switzerland), University of Pavia (Italy), University of Trento (Italy), VTT (Finland).

The activities of the IIM-TS project are:

1. identification of the state-of-the-art applications, methods, and algorithms for detection and analysis of meaningful features (information) in image time series obtained by SAR and optical sensors (e.g., global and local scale classification, change detection, etc.); and
2. selection, design and implementation of prototype tools for inclusion in the KEO system.

Role of Iguassu in IIM-TS

The image information mining, especially, when extended for multi-temporal analysis, deals with large amount of data and requires high-performance computing methodologies such as algorithm parallelization together with employing of the computer clusters or using distributed GRID computing. The role of ISS in IIM-TS project is the assessment and trade-off of different options of algorithm parallelisation on different SW and HW platforms such as the existing KEO HW infrastructure, GRID (at the middleware level), or G-pod (GRID Processing On Demand) ESA GRID infrastructure highly customised for Earth Observation applications [3]. The activities include interconnection of the KEO system with the considered computing platforms, if possible using, the Web Service SOAP interface, and preparations of prototypes of the sample parallelised algorithms selected by the consortium. The outputs of the study shall provide information for decision of the optimal parallelization strategy within the given constrains (HW, KEO architecture, MPI etc.) to be adopted by KEO developers.

[1] http://earth.esa.int/rtd/Projects/

[2] http://earth.esa.int/rtd/Projects/KEO/

[3] http://gpod.eo.esa.int/