Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Classification of Maritime Vessels using Convolutional Neural Networks

 
: Anneken, Mathias; Strenger, Moritz; Robert, Sebastian; Beyerer, Jürgen

:
Fulltext urn:nbn:de:0011-n-6217597 (2.2 MByte PDF)
MD5 Fingerprint: 6de5243328d59199d1ca4ac799bceb1d
Created on: 28.1.2021


Christ, Andreas (Ed.):
Artificial Intelligence. Research Impact on Key Industries. The Upper-Rhine Artificial Intelligence Symposium, UR-AI 2020 : Collection of Accepted Papers of the Canceled Symposium, Karlsruhe, 13th May 2020
Offenburg: Hochschule für Technik, Wirtschaft und Medien, 2020
ISBN: 978-3-943301-28-1 (Print)
ISBN: 978-3-943301-29-8 (Online)
pp.103-114
The Upper-Rhine Artificial Intelligence Symposium (UR-AI) <2020, Karlsruhe/cancelled>
English
Conference Paper, Electronic Publication
Fraunhofer IOSB ()
residual neural network; time series classification; Convolutional Neural Networks; maritime domain; ship classification

Abstract
Due to a steady increase in traffic at sea, the need for support in surveillance task is growing for coast guards and other law enforcement units all over the world. An important cornerstone is a reliable vessel classification, which can be used for detecting criminal activities like illegal, unreported and unregulated fishing or smuggling operations. As many ships are required to transmit their position by using the automatic identification system (AIS), it is possible to generate a large dataset containing information on the world wide traffic. This dataset is used for implementing deep neural networks based on residual neural networks for classifying the most common ship types based on their movement patterns and geographical features. This method is able to reach a competitive result. Further, the results show the effectiveness of residual networks in time-series classification. Due to a steady increase in traffic at sea, the need for support in surveillance task is growing for coast guards and other law enforcement units all over the world. An important cornerstone is a reliable vessel classification, which can be used for detecting criminal activities like illegal, unreported and unregulated fishing or smuggling operations. As many ships are required to transmit their position by using the automatic identification system (AIS), it is possible to generate a large dataset containing information on the world wide traffic. This dataset is used for implementing deep neural networks based on residual neural networks for classifying the most common ship types based on their movement patterns and geographical features. This method is able to reach a competitive result. Further, the results show the effectiveness of residual networks in time-series classification.

: http://publica.fraunhofer.de/documents/N-621759.html