A CAE-Based Method for Anomaly Detection in Ship Trajectories

Ship trajectory anomaly detection aims to identify abnormal behaviors from a large amount of trajectory data, which is crucial for revealing or preventing potential maritime traffic risks. This study develops a novel trajectory anomaly detection method based on convolutional autoencoder (CAE). It first converts the ship's motion trajectory into the form of image vectors, then uses convolutional autoencoder for modeling and training. The training target of the autoencoder (AE) is to maximize the reconstruction of the original trajectory data in the decoding process. After training is completed, the model can effectively identify trajectory anomalies by comparing the differences between the input trajectory and the reconstructed trajectory by the model. The advantages of this method are that it only needs to compare the differences between the original trajectory and the reconstructed trajectory, which significantly reduces the computational cost. In addition, the model allows automatically setting the normal and abnormal trajectory distance threshold by analyzing the proportion of abnormal trajectories in the training data set, avoiding the trouble of manually setting thresholds, making the model easier to apply in various practical scenarios. Performance tests show that the model shows excellent convergence performance during training and achieves a high standard of over 90% in accuracy.

I. Introduction

With the increasing demand for various sensor devices and location-based services, Automatic Identification Systems (AIS) are widely used in the maritime industry. Traditional maritime activities often involve human intervention in the operation and monitoring of vehicles, which can be challenging to adapt to the increasingly complex maritime traffic environment [1] and may lead to operator fatigue and operational errors. Nowadays, the exponential growth of maritime data is facilitated by various sensors and communication devices [2]. Automatic Identification Systems can enhance or replace manual monitoring and surveillance in ports. Analyzing AIS data using machine learning and deep learning algorithms can help accurately model ship trajectories and provide decision support for assisting human navigation detection systems.