ESA LTDP PFA - Product Feature Extraction and Analysis
Fomferra, Norman¹; Bovolo, Francesca²; Bruzzone, Lorenzo²; Danne, Olaf¹; Demir, Begün²; Iapaolo, Michele³; Iha, Rajesh⁴; Lu, Jun⁴; Marchetti, Pier Giorgio³; Quast, Ralf¹; Veci, Luis^4
1Brockmann Consult GmbH, GERMANY; ²University of Trento, ITALY; ³ESA ESRIN, ITALY; ⁴Array Systems Computing Inc, CANADA

Even though the use and exploitation of Earth Observation data is increasing, the users of the data still do not manage to fully exploit the existing satellite data archives. In particular, current data selection criteria are still mainly based on time and geographic location, thus leaving to the user and/or scientist the burden to contextualize specific EO product features, which may or may not be present in the selected datasets.

The aim of the Product Feature Extraction (PFA) project is to demonstrate that feature extraction and analysis of EO data makes it possible to exploit the large volumes of various types of satellite data more efficiently, and through this to foster the scientific analysis of mission wide datasets. The PFA project addresses emerging methods and tools for data product features and information extraction, in view of possible implementations for enriching data description and easing the use of archived optical and SAR data. For this purpose, the PFA team is elaborating and implementing a number of EO data selection and exploitation scenarios, whose usefulness is demonstrated by means of real-life applications. Within the project state-of-the-art image information mining methods will be developed taking into account sensor synergies, data fusion and methods developed in other domains such as medical, microscopic, astronomic. For the demonstration system, the PFA project will make use of the full mission datasets of ASAR, ERS, Landsat TM, MERIS and AATSR in order to compare its performance with respect to the data streams generated by the upcoming Sentinel 1, 2 and 3 missions.

The implementation of the algorithms will be done using ESA's BEAM (for optical data) and NEST (for SAR data) toolboxes. A high-throughput EO data processing system based on Apache Hadoop will be used for the extremely demanding feature extraction task. The system to be developed targets at later integration with existing ESA user services for data delivery and processing such as ngEO and G-POD.