Abstract
This research contributes to the literature with a methodology for the extraction of standard naval routes from AIS data (data from international naval Automatic Identification System). A fundamental objective of our work is to sup- port environmental surveillance of maritime activities, including the detection of anomalous ship behavior and safe&rescue operations, both involving a need to define the expected behavior of ships in terms of standard routes. Previous re- search mostly focuses on the comparison of the performance of different statisti- cal and machine learning techniques. In our work, we draw from previous re- search the best performing ML approach (unsupervised) and we focus on the de- sign of an end-to-end data processing pipeline that combines ML techniques with data pre-processing rules and geographical information with the goal of making data usable. Our methodology has been tested on a 3-year time series of AIS data, with a focus on the Arctic region. Results show how standard routes can be ef- fectively extracted with good precision owing to the robust sequence of our meth- odological steps. Our algorithm has identified roughly 400 standard routes in the Arctic region, starting from a total of over 140 million AIS messages. Results show how working on data cleaning, accurately calculating distance among routes, and applying two phases of clustering on different sets of features are crucial for the dependability of results. Further work will be conducted in the ARCOS project to build a complete, longitudinal dataset of standard routes to be released as open data.
Recommended Citation
Corsi, Marco; Francalanci, Chiara; Giacomazzi, Paolo; Ravanelli, Paolo; and Tilia, Simone, "Extracting standard naval routes from AIS data: Insights from an In-Depth Analysis of the Arctic Region" (2023). ITAIS 2023 Proceedings. 19.
https://aisel.aisnet.org/itais2023/19