Encounter Detection Using Visual Analytics to Improve Maritime Domain Awareness

A visual analytics process to detect encounters between vessels from ship position data is developed in this thesis. An archive of historical position records is pre-processed and filtered to provide input for an encounter detection algorithm. The algorithm arranges the position records into a set of sorted lists SSL so that only a minimum number of records need to be compared. The algorithm performs a single sweep over the record set to arrange it into a SSL and simultaneously find the encounters. To avoid problems due to discrete sampling, an interpolation of the data is performed when the sampling is too sparse. To accommodate large data sets, a divide-and-conquer approach using a sliding spatial window is developed. In post-processing, the elementary encounters are grouped into composite encounters by collecting elementary encounters occurring between the same vessels. Additionally, the composite encounters are input into a visual analytics tool where each composite encounter is represented as a layer on a map. Patterns of life analysis and investigations of potential anomalous activity are performed by zooming in on encounter areas of interest. The development of a visual analytics process to identify vessels of interest is the significant result of this thesis.