Detecting Anomalous Ship Movements in Indonesian Seas Using Convolutional Neural Networks

Detecting unusual ship movements is a crucial feature of maritime surveillance, particularly in Indonesian waters, where illegal fishing, unauthorized resource exploitation, drifting ships, and unauthorized navigation pose significant threats to safety and security. This research proposes a Convolutional Neural Network (CNN)-based methodology for categorizing ship movement behaviors into two classifications: drifting and non-drifting. The dataset has 79,200 image-based samples, uniformly divided between the two categories. The proposed model is trained and tested using accuracy, recall, precision, F-score performance metrics. The experiment shows that the resulting model successfully classifies the movement of the ship well. This is evidenced by a testing accuracy of 0.98, a precision of 99%, a recall of 95%, an F-score of 97%, indicating that the CNN was highly accurate and robust, suggesting it could be utilized in real-time maritime anomaly detection systems.