Abstract
Background Georeferencing sensors are available in people, animals, or objects to generate trajectory data for relevant civilian and military applications. A common task in these applications is detecting and recognizing organized group movement patterns. Methods based on Shannon Entropy and Spectral Clustering are two widely used general approaches, however, they fail to detect organized group movements for various motion patterns.Methods To address these limitations, this article proposes a Network Inference-based approach followed by clustered data analysis and evaluation of cluster quality using the DBSCAN and Silhouette Coefficient algorithms.Results In our experiments, we compared our approach against Shannon Entropy and Spectral Clustering on simulated and real data, using the organization index median in a run, as the metric. Specifically, the proposed method had superior performance in three simulated scenarios, which were Ants, Wolf Sheep Predation, and Flocking, and in two real data sets, which are Fish School Trajectories and PETS09-S2L1.Conclusion Our approach presents a promising solution for identifying organized group movements, critical for emergency decision-making and resource optimization in practical applications.