Evaluating Neural Networks for Early Maritime Threat Detection

TL;DR

This paper evaluates neural network models for classifying maritime boat trajectories to detect threats, demonstrating superior accuracy over traditional entropy-based methods, especially with full trajectory data.

Contribution

It introduces neural network approaches for maritime threat detection, outperforming entropy-based clustering, and explores data augmentation and normalization techniques for robustness.

Findings

Deep networks achieve up to 100% accuracy on full trajectories.

Accuracy degrades gracefully with fewer time steps.

Neural networks outperform entropy-based clustering methods.

Abstract

We consider the task of classifying trajectories of boat activities as a proxy for assessing maritime threats. Previous approaches have considered entropy-based metrics for clustering boat activity into three broad categories: random walk, following, and chasing. Here, we comprehensively assess the accuracy of neural network-based approaches as alternatives to entropy-based clustering. We train four neural network models and compare them to shallow learning using synthetic data. We also investigate the accuracy of models as time steps increase and with and without rotated data. To improve test-time robustness, we normalize trajectories and perform rotation-based data augmentation. Our results show that deep networks can achieve a test-set accuracy of up to 100% on a full trajectory, with graceful degradation as the number of time steps decreases, outperforming entropy-based clustering.