NUK - logo
E-viri
Celotno besedilo
  • A First Approach to the Automatic Detection of Marine Litter in SAR Images Using Artificial Intelligence
    Savastano, Salvatore; Cester, Iva; Perpinya, Marti; Romero, Laia

    2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, 2021-July-11
    Conference Proceeding

    Remote sensing, and particularly Sinthetic Aperture Radar (SAR) is a potential tool to monitor the presence of plastic on the ocean surface, and AI algorithms can learn to find patterns in the satellite signals that lead to authomatic detection of plastic patches. The main challenge to train AI algorithms for this purpose is the lack of insitu ground truthdata that allow to reliably label the images. This paper describes a first pilot of binary pixel classification (plastic vs. non-plastic) in SAR imagery using Random Forest, GaussianNaïve Bayes and Support Vector Machines classifiers. It also describes the strategy used to generate ground truth data using Sentinel-1 and Sentinel-2 in the Balearic Islands.

Vnos na polico