Trajectory analysis is one of the most actively researched areas of spatio-temporal databases. Exploring and analysing large datasets of movement data has become a vital part of research in many disciplines and decision-making fields. The major challenge involved during the analysis process of trajectory data is to visualize, understand and extract meaningful patterns (Adrienko & Adrienko, 2011) out of millions of locations collected from Automatic Identification Systems (AIS) data points. AIS datasets are used in the maritime industry to assist in tracking and monitoring vessel movements. The ultimate aim of the study is to understand the characteristics of different types (Dodge, Weibel, & Forootan, 2009) of vessels using AIS movement data. The intention of the study is also to outline challenges encountered during this thesis and describe approaches taken to overcome them. AIS movement datasets are voluminous and are coded via a complex standard. Therefore, to conduct analysis on raw data to trajectory involved a two-phased methodological process. The first phase focused on development of a decoder to extract significant information from the raw data. The information extracted from movement data was then utilized to perform knowledge discovery in regard to dynamic objects. The second phase centred on trajectory analysis utilizing proposed spatio-temporal approach and clustering techniques. Each phase accounted for a part of the contribution made towards this thesis.

Phase 1 primarily focused on handling the large dataset and development of a decoder. Given a large dataset (2GB - 30 Million rows) of spatio-temporal movement observations, the goal was to perform a segmentation analysis into sets of certain period of data. The objective of segmentation was to handle the datasets into manageable sizes to overcome the main memory issue. The purpose of the decoder is to decode the raw data to read the information signalled from the vessels. The decoder is developed in VB.net that could extract information such as Maritime Mobile Service Identity (MMSI), latitude, longitude, speed over ground (SOG) and time. The developed decoder has the capability to intake raw data in text files and output the decoded dataset in text files. As the decoded dataset is made available, it is imported to an Excel file. The most important challenge faced was clean the decoded data from noise, outliers and duplications. A filtering technique was utilized in the process of cleaning. Subsequently, the dataset is well prepared for the next phase of conducting trajectory analysis within the Geographical Information Systems (GIS). In Phase 2, trajectory analysis was conducted using a spatio-temporal approach that included methods such as projecting trajectories on ArcGIS, plotting velocity trend graphs (Guo, Liu, & Jin, 2010) and identifying clusters based on stop or start movements. Exploring and analysing large datasets of movement data is a central aspect of this research study. The research focused on methodological research aimed extracting the spatio-temporal information and patterns of moving objects.

The experimental results indicate that the proposed methods can be successfully applied to perform trajectory analysis on provided movement datasets of dynamic vessels.