Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Maritime Anomaly Detection with Dynamic Potential Field Grids
Blekinge Institute of Technology, Faculty of Computing, Department of Creative Technologies.
Blekinge Institute of Technology, Faculty of Computing, Department of Creative Technologies.
2015 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Context. Safety and security in the maritime domain is an important area and systems designed to aid in detecting illegal or dangerous situations are desired. With the advent of Automatic Identification Systems (AIS), information about marine traffic is in abundance and the challenge is in finding events of interest; to which a common approach is to find vessels behaving anomalously.

Objectives. This study extends previous work by Osekowska et al. who used potential fields to condensate AIS data into a unified pattern. The challenge in this method is the configuration of grid precision. Through the use of quadtrees, this thesis presents a new implementation that intends to alleviate this problem by allowing dynamic grid resolution.

Methods. Experiments are conducted to compare the new implementation to the existing one. Both algorithms are tested in areas with dense, sparse and mixed traffic density. Additionally, the implementations are validated against real anomaly data and a set of traffic data containing no confirmed anomalies.

Results. While the new algorithm does not work exactly like the existing implementation in dense and sparse areas, it is shown to be able to detect 8/9 real anomalies while detecting 20.1\% false positives. The existing implementation either finds 5/9 of the real anomalies and 8.3\% false positives with a large cell size or 7/9 real anomalies and 44.4\% false positives with a small cell size.

Conclusions. The new algorithm finds more real anomalies, especially compared to when using a large cell size. It finds fewer false positives than when using small cells, but not quite as few as with large cells. The anomaly missed by the new algorithm is likely due to the fact that traffic patterns have changed after the training data was collected, and it is possible that it would have been found with more up-to-date training data. This thesis is a pre-study of the presented method and provides incentive for future investigation. 

Place, publisher, year, edition, pages
2015. , 46 p.
Keyword [en]
potential fields, maritime anomaly detection, quadtrees, supervised learning
National Category
Computer Systems Computer Science
Identifiers
URN: urn:nbn:se:bth-10394OAI: oai:DiVA.org:bth-10394DiVA: diva2:839528
Subject / course
DV2524 Degree Project in Computer Science for Engineers
Educational program
PAACI Master of Science in Game and Software Engineering
Supervisors
Examiners
Available from: 2015-08-05 Created: 2015-07-02 Last updated: 2016-10-05Bibliographically approved

Open Access in DiVA

No full text

By organisation
Department of Creative Technologies
Computer SystemsComputer Science

Search outside of DiVA

GoogleGoogle Scholar

Total: 758 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf