A Deep Learning Streaming Methodology for Trajectory Classification

Abstract

Due to the vast amount of available tracking sensors in recent years, high-frequency and high-volume streams of data are generated every day. The maritime domain is no different as all larger vessels are obliged to be equipped with a vessel tracking system that transmits their location periodically. Consequently, automated methodologies able to extract meaningful information from high-frequency, large volumes of vessel tracking data need to be developed. The automatic identification of vessel mobility patterns from such data in real time is of utmost importance since it can reveal abnormal or illegal vessel activities in due time. Therefore, in this work, we present a novel approach that transforms streaming vessel trajectory patterns into images and employs deep learning algorithms to accurately classify vessel activities in near real time tackling the Big Data challenges of volume and velocity. Two real-world data sets collected from terrestrial, vessel-tracking receivers were used to evaluate the proposed methodology in terms of both classification and streaming execution performance. Experimental results demonstrated that the vessel activity classification performance can reach an accuracy of over 96% while achieving sub-second latencies in streaming execution performance.

A Comparison of Trajectory Compression Algorithms Over AIS Data

“…This denotes that the HD compression algorithm is less volatile and it is an indicator that it is more suitable for datasets in the maritime domain. This is further confirmed by research conducted in [17] where the proposed trajectory classification algorithm achieves a better classification performance when the HD is used in trajectories compared to the rest of the algorithms.…”

“…To evaluate the amount of information loss that is posed by each compression algorithm, trajectory similarity measures were chosen (see Section IV) that are able to measure the distance between the original trajectory and the compressed one in terms of shape. The shape of the trajectory and the pattern that it forms can play an important role in the identification of illegal activities [3], [17]. Therefore, for each compression algorithm, we measured the distances D between each trajectory and its compressed counterpart and calculated the mean distance and the standard deviation of distance for the entirety of the dataset.…”

“…The purpose of a similarity measure is to obtain a quantitative measure between any two trajectories, thus to identify to what extent two objects are similar. Most of the well-established similarity measures selected in this work, consider the shape of the trajectory which consist a characteristic of utmost importance in the maritime domain and can provide insights about vessels' behavior [3], [17].…”

PSOTSC: A Global-Oriented Trajectory Segmentation and Compression Algorithm Based on Swarm Intelligence

“…The segmentation algorith in TRACLUS will be affected by trajectory scaling, and different scaling ratios will aff the segmentation accuracy. The scaling ratio in the experiment [1,2,3,5,10,15,20,25,30] TRACLUS r  . The segmentation accuracy of the TCSS algorithm affected by the angle threshold setting, and the angle threshold setting in the experime…”

“…Trajectory data contains rich information of target space-time characteristics. Through data mining and depth analysis, we can find high-value information, such as target activity law, behavior characteristics, interest habits, abnormal changes [1][2][3][4][5][6], etc. However, the rapid growth of trajectory data also brings many challenges to the data service based on the target space-time location, including the increase in data transmission load, the pressure of data storage, the reduction of data query efficiency and the decline of data analysis performance [7].…”