A Framework for the Detection of Search and Rescue Patterns Using Shapelet Classification

Abstract

The problem of unmanned supervision of maritime areas has attracted the interest of researchers for the last few years, mainly thanks to the advances in vessel monitoring that the Automatic Identification System (AIS) has brought. Several frameworks and algorithms have been proposed for the management of vessel trajectory data, which focus on data compression, data clustering, classification and visualization, offering a wide variety of solutions from vessel monitoring to automatic detection of complex events. This work builds on our previous work in the topic of automatic detection of Search and Rescue (SAR) missions, by developing and evaluating a methodology for classifying the trajectories of vessels that possibly participate in such missions. The proposed solution takes advantage of a synthetic trajectory generator and a classifier that combines a genetic algorithm (GENDIS) for the extraction of informative shapelets from training data and a transformation to the shapelets’ feature space. Using the generator and several SAR patterns that are formally described in naval operations bibliography, it generates a synthetic dataset that is used to train the classifier. Evaluation on both synthetic and real data has very promising results and helped us to identify vessel SAR maneuvers without putting any effort into manual annotation.

A Deep Learning Streaming Methodology for Trajectory Classification

“…The main disadvantage is that these features cannot be used to classify other mobility patterns, and they are only suitable for the fishing patterns in the study. Kapadais et al [2] proposed a time-series shapelet classification technique to identify SAR patterns. The main drawback in such techniques is that they require data to be available at fixed intervals (e.g., hourly, weekly, monthly, yearly), a characteristic that is not present in AIS messages (see Section 4.1).…”

“…They tested their approach on two surveillance data sets, MIT car [26] and T15 [27], yielding promising results. Kapadais et al [2] treated the problem of trajectory classification as a time-series or shapelet classification task. The main problem though with time-series classification is that values or features such as the speed of the vessel need to be reported at fixed time intervals, which is not the case of the AIS protocol -vessels traveling at higher speeds report their position more frequently (the same thing happens when the change of rate of a turn is also higher).…”

“…On the other hand, trajectory clustering approaches are often employed to form groups of AIS positions with similar spatiotemporal behaviors, uncovering behaviors that are harder to predefine. Although there is an abundance of studies in the literature regarding offline trajectory classification and clustering [1][2][3][4][5], fewer works have focused on steam processing of events in the maritime domain [6][7][8][9][10]. Event processing methodologies are faced with significant challenges when employed on streaming data where the requirements for such applications demand low memory consumption and decreased latencies.…”

“…marinetraffic.com/hc/en-us/articles/217631867-How-often-do-the-positions-of-thevessels-get-updated-on-MarineTraffic-, accessed on 7 April 2021). This is in contrast to time-series methodologies [2] that are inherently unsuitable for such tasks and require data points at fixed time points; • Trajectory classification approaches found in the literature [4,10,14,15] require a preprocessing step such as the understanding and analysis of data and the selection of features suitable only for the mobility patterns to be classified. This means that features selected for a certain mobility pattern cannot be applied to other patterns as well [14].…”

Explainable Artificial Intelligence on Smart Human Mobility: A Comparative Study Approach

“…On human mobility context some XAI projects have been studied. For example, [8] focuses on the topic of automatic detection of Search and Rescue (SAR) missions, by developing and evaluating a methodology for classifying the trajectories of vessels that possibly participate in such missions. Luca et al provides a taxonomy of mobility tasks where a discussion on the challenges related to each task and how deep learning may overcome the limitations of traditional models and a description of the most relevant solutions to the mobility tasks are described [10].…”