CIRFA seminar 23 Jan. 14:00 – Anja Strømme and Ding Tao

CIRFA welcomes you to a seminar where Anja Strømme from the Norwegian Space Centre will talk about Norways involvement in the Copernicus program and Ding Tao from Fudan University will present work on machine learning and sea ice.

Date: Tuesday 23 January 2018 at 14h00-15h00

Venue: CIRFA, SIVA Innovation Centre (Forskningsparken), Tromsø [map]

Online: https://connect.uninett.no/cirfa-seminar/

Norwegian Space Centre, ESA, Copernicus

Anja Strømme, Senior Advisor, Norwegian Space Centre

The presentation will introduce the Norwegian Space Centre and its activities towards ESA and the Copernicus program, including the national ground segment.

 

 

 

 

Sea Ice Image Generation and Analysis based on Arctic SAR Data and Generative Adversarial Networks

Ding Tao, Young Research Associate Professor, School of Information Science and Technology, Fudan University

This work discusses the potential applications of deep learning method in synthetic aperture radar (SAR) image analysis for ice-infested Arctic waters. For such complicated conditions, the information of various maritime targets, e.g., changing sea ice, icebergs, ships, man-made platforms, and oil spills, becomes increasingly valuable. Space-borne SAR can provide continuous wide-swath monitoring of the Arctic region in all weather conditions independent of daylight. Nowadays, there are many commercial SAR missions, including Sentinel-1 (ESA), ALOS-2 (Japan), RADARSAT-2 (Canada), etc., offering large amount of observing data. In addition, recent deep learning methods with the ability to learn geometrical information of various images have received lots of attentions. Such methods supported by the available SAR data have great potential for Arctic characteristics studies, and in particular, for information extraction of different targets within the high-latitude region. In this study, we utilize the Generative Adversarial Networks (GANs), where the generative and discriminative networks are trained simultaneously and compete with each other. Thus, two unsupervised schemes for ROI clustering and target discrimination are proposed for ice-infested Arctic waters. Experiments are conducted on real RADARSAT-2 HH polarization MLI SAR images.

 

Print Friendly, PDF & Email

Comments are closed.