A novel approach exploiting properties of convolutional neural networks for vessel movement anomaly detection and classification

Abstract

The article concerns the automation of vessel movement anomaly detection for maritime and coastal traffic safety services. Deep Learning techniques, specifically Convolutional Neural Networks (CNNs), were used to solve this problem. Three variants of the datasets, containing samples of vessel traffic routes in relation to the prohibited area in the form of a grayscale image, were generated. 1458 convolutional neural networks with different structures were trained to find the best structure to classify anomalies. The influence of various parameters of network structures on the overall accuracy of classification was examined. For the best networks, class prediction rates were examined. Activations of selected convolutional layers were studied and visualized to present how the network works in a friendly and understandable way. The best convolutional neural network for detecting vessel movement anomalies has been proposed. The proposed CNN is compared with multiple baseline algorithms trained on the same dataset.

Keywords: Anomaly detection; Convolutional neutral networks; Deep learning; Radar datasets; Vessel movement anomalies.