Loading...
 
Print

ECEO

Introduction and Objectives

The E-Collaboration for Earth Observation (E-CEO) project addresses the technologies and architectures needed to provide a collaborative research platform for automating data mining and information extraction experiments.
The platform serves for the implementation of open contests focusing on Information Extraction for EO applications. The possibility to implement – through the contests – different approaches and algorithms in a common Software Environment (provided by the collaborative platform) facilitates the comparison and the evaluation between different methodologies, which is one of the main requirements requested by scientific experts who develop algorithms in the EO field.
The E-CEO main objective is to provide a collaborative research platform for:

  1. Automation of data mining and information extraction experiments;
  2. Generation of reproducible results that can be easily shared;
  3. Addressing specific scientific challenges and tackling new research problems in a “parallel and collaborative way”.

Expected Results

One of the main results is the execution of real contests in the field of Earth Observation. In particular:

Atmospheric Correction Data Challenge
This Challenge is based around ocean colour AC and the use of auxiliary information that’s needed when processing the data from Level 1 (Top of Atmosphere, TOA, calibrated radiances/reflectances) to Level 2 products (Bottom of Atmosphere, BOA, calibrated radiances/reflectance and derived products).
Scientific researchers often accept the auxiliary inputs that they’re provided with and/or use the climatological data that accompanies the processing software; commonly because it can be difficult to obtain multiple data sources and convert them into a format the AC processing software accepts. Therefore, the Challenge will be used to compare various ocean colour AC approaches and in the process also study the uncertainties associated with using different meteorological auxiliary products for the processing of Medium Resolution Imaging Spectrometer (MERIS) imagery i.e. the sensitivity of different AC assumptions.
Therefore, in summary, a challenge to run different AC models / code on the E-CEO platform with varying meteorological auxiliary information can provide an insight into both the models and variations caused by (sensitivity to) the auxiliary information.
To encourage community involvement, the Challenge is being advertised through the European Geosciences Union (EGU) General Assembly 2014 (Vienna, Austria) from 27 April – 02 May 2014 as a poster presentation by Dr Lavender within the ‘Ocean Remote Sensing’ session, existing ESA projects / activities (such as the OC_CCI and MERIS QWG) and through the mailing lists of organisations such as IOCCG. The aims is to have submissions from new/novel approaches that can be tested alongside more accepted approaches
The plan is for the Challenge to have the deadline for the submission of code / algorithms as the 31 July 2014, with the evaluation then lasting around 2 months so that the preliminary results are known at the end of September 2014. This would allow the preliminary results to be presented at an E-CEO project review meeting in October 2014 with distribution of the results to the participants (in an anonymised manner unless otherwise agreed so, see Section 3.3) and public announcement of the ‘winner’ following soon after. Depending on the final outcome, a peer-reviewed publication is also envisaged with participants being invited to be co-authors.
AC contest brochure

Cloud4SAR Interferometry Data Challenge
Increasingly, there is interest for algorithms and computing platforms capable of delivering InSAR solutions in an easy, efficient and widely accessible way. SAR Interferometry (InSAR) is a well-established technique to spatially and temporally detect ground displacements at millimetre scale. The huge archives of Synthetic Aperture Radar data acquired during the last 20 years by National Space Agencies and ESA, have played a significant role in making InSAR particularly suitable not only for ground deformation analysis but also for natural hazard monitoring and risk management. Increasingly, there is interest for algorithms and computing platforms capable of delivering InSAR solutions in an easy, efficient and widely accessible way for several scientific, civil protection and commercial communities. The emerging Cloud Computing paradigm offers an integrated approach to address all these needs.
In this scenario, we set up a Data Challenge to foster the deployment of InSAR processing chains within a Cloud platform. While a large variety of InSAR processing software is available, it requires a high level of expertise and a complex user interaction to be run. Computing a co-seismic interferogram or a 20-years deformation time series on a volcanic area are not easy tasks to be performed in a fully unsupervised way or in very short time (hours or less). Benefiting from ESA's e-Collaboration platform, participants can optimise algorithms on a Virtual Sandbox environment without being expert programmers, and compute results on high performing Cloud platforms.
This contest requires to solve a relatively easy InSAR problem by trying to maximize the exploitation of the processing capabilities provided by a Cloud Computing infrastructure.
The proposed challenge offers two different frameworks, each dedicated to participants with different skills, identified as Beginners and Experts. For both of them, the contest mainly resides in the degree of automation of the deployed algorithms, no matter which one is used, as well as in the capability of taking effective benefit from a parallel computing environment.

Beginners in the Cloud4SAR contest

Problem given a set of SAR data acquired over an active seismic region, produce the earthquake co-seismic displacement map using unsupervised methods

Experts in the Cloud4SAR contest

Problem starting from SLC SAR images acquired on a volcanic area, generate InSAR displacement time series and the relevant mean velocity map
Winners get publication on the ESA website

Cloud4SAR contest brochure

Contractors

The E-CEO consortium is made of four companies and one research institution:

  • Terradue UK Ltd (UK) - leads the consortium bringing its unique expertise on the G-POD System, CIOP, SSEP and SSOi projects for the design and implementation of the e-Collaboration for EO platform;
  • Science and Technology AS (NO) - participates in the e-Collaboration for EO platform requirements specifications, technological analysis and validation and in the contest #2 definition;
  • CNR-IREA (IT) - leads the Contest #1 Interferometry through it definition, implementation and evaluation;
  • Pixalytics (UK) - is responsible for the Contest #3 definition
  • UNINOVA (PT) - leads the Evaluation and Ranking definition and process

Project Deliverables





Contributors to this page: Michele Iapaolo

.

Page last modified on Monday 02 of May 2016 11:02:00 CEST by Michele Iapaolo.