Lieu : salle C406, Bâtiment Cournot, ENS Paris-Saclay, campus de Cachan
Date et horaire : lundi 27 janvier 2020, 12h00-13h00
3-D reconstruction from tomographic SAR data: From low-to high-level information processing
par Olivier D’Hondt (Technische Universität Berlin)
Tomographic SAR (TomoSAR) is a powerful technique that allows the 3-D reconstruction of objects lying on the surface of the Earth. It relies on a system called Synthetic Aperture Radar that sends and receives electromagnetic waves in the microwave domain and produces 2-D complex-valued images. As opposed to optical sensors, SAR is not affected by weather conditions and operates by day and night. By combining the phase of several images acquired from different points of view, it is possible to achieve 3-D imaging of man-made objects such as buildings, but also natural targets such as trees. Because at certain frequencies the wave penetrates the vegetation, the topographic structure of the ground beneath the canopy can be recovered, which would be impossible with an optical sensor.
This talk will cover the basic principles of TomoSAR as well as more advanced research topics in the field of image processing and machine learning carried on by the Computer Vision and Remote Sensing team at TU-Berlin. An emphasis will be put on some research challenges that are currently faced by the SAR community as well as potential applications and future satellite missions.