OSIRIDE Short Description

Content-based image retrieval (CBIR) techniques are successfully used in several domains like multimedia data bases search, crime prevention or medical diagnosis. Due to particularities of the Earth Observation (EO) data imagery, there are, though, still, many open issues needed to be addressed and to be solved in order to put in place efficient and reliable retrieval systems. Adapting CBIR for the EO domain is of main importance to fully exploit the next generation of satellite images.

The main objective of the OSIRIDE project is to develop, implement and integrate tools for Earth Observation Content Based Image Retrieval into a powerful and ready-to-use Open-Source platform.

OSIRIDE tries to expand the data mining perspective into Big Data analytics. A modular architecture integrates several methods and algorithms to discover hidden patterns inside the data matching particular characteristics, to visualize the data given a feature space in order to guide a query, to learn semantic dependencies based on a relevance feedback received from the user and to link the physical parameters (EO data) with semantic labels (value-added products). Even if the access to the system is made via a web platform, the user is not allowed to add new data into the repository. However, he is allowed to perform complex queries and train the system through an active learning process.

Project Details