Joint ONERA – CNAM Workshop
(ONERA, Palaiseau, April 5th, 2018, 10:00 – 15:00)
Mihai Datcu presented an overview of the particularities of Artificial Intelligence (AI) in the fields of Earth Observation (EO). The focus was on the methods of Deep Learning and their specificities for multispectral and Synthetic Aperture Radar. Have been presented various relevant examples, which differentiate the methods in EO from other areas as Computer Vision. The conclusion emphasizing the main research directions of importance in EO both in theory and in practical applications.
The lecture was followed by the presentation of the recent results obtained at ONERA in the field and live discussion and debate on the perspectives of AI in EO.
AI4SAR: Artificial Intelligence for Synthetic Aperture Radar
Talk by Mihai Datcu
(Centrale-Supélec, 3 rue Joliot-Curie, 91190 Gif-sur-Yvette: Salle du Conseil, Laboratoire des Signaux et Systèmes (L2S), Bâtiment Breguet, 11:00-12:30, May 4th, 2018; Café/croissants from 10:30)
The challenges of the Synthetic Aperture Radar (SAR) image formation principles, the high data volume and the very high acquisition rate stimulated from the very beginning the elaborations of sophisticated techniques. Meanwhile the SAR technologies have immensely evolved. The state of the art sensors deliver widely different imaging modes, and have made considerable progress in spatial and radiometric resolution, target acquisition strategies, or geographical coverage and data rates. Generally imaging sensors generate an isomorphic representation of the observed scene. This is not the case for SAR, the observations are a doppelganger of the scattered field, an indirect signature of the imaged object. This spots the load of SAR image understanding, and the outmost challenge of Big SAR Data Science, as a new and particular challenge of Machine Learning (ML) and Artificial Intelligence (AI). The presentation reviews and analyses the new approaches of SAR imaging leveraging the recent advances in physical process based ML and AI methods and signal processing. These is leading to Computational Imaging paradigms where intelligence is the analytical component of the end-to-end sensor and Data Science chain design. A particular focus is on the scientific methods of Deep Learning and an information theoretical model of the SAR information extraction process.
Earth Observation Data Science: particularities and challenges
Talk by Mihai Datcu
(CNAM, Amphitheatre George Friedmann, 10:30 – 12:00, March 9, 2018)
The volume and variety of valuable Earth Observation (EO) images as well as non-EO related data is rapidly growing. The open free data access becomes widespread and has an enormous scientific and socio-economic relevance. EO images are acquired by sensors on satellite, suborbital or airborne platforms. They extend the observation beyond the visual information, gathering physical parameters of the observed scenes in a broad electromagnetic spectrum. The sensed information depends largely on the imaging geometry, orbit, illumination and other specific parameters of the space instruments. Typical EO systems can be classified into optical or radar instruments. During the last years, both types of sensors deliver widely different images, and both have made considerable progress in spatial and radiometric resolution, image acquisition strategies, and data rates.
Therefore, the domain of EO Data Science spans from mathematical models for the satellite orbit, the physics of electromagnetic propagation and scattering, signal processing, machine learning, AI and knowledge representation. The new specific EO methods to Data Analytics and Data Mining are designed to leverage advances in physical parameters extraction and leads to AI paradigms where learning is a critical non-traditional element joining the overall EO system parameters to achieve the optimal information extraction. This involves the inclusion of machine or deep learning and reasoning tools in the EO data processing chains together with supplementary external information to analyze target characteristics, to perform classification, to search for similarities or extract semantic entities. Information-theoretic approaches are the fundament for modeling the end-2-end system behavior and analyze its performance and bounds.