Détection de Comportements Anormaux dans des Trajectoires de Navires avec One-Class SVM et Dynamic Time Warping

Anomaly detection in the field of maritime surveillance is of major importance for the safety of ships and nations. Numerous anomaly detection algorithms are available in the state-of-the-art but they are not necessarily adapted to the analysis of time series such as ship trajectories, in particular when these trajectories have different lengths. This article studies an algorithm allowing the One-Class SVM method to be adapted to time series associated with ship trajectories by means of a kernel based on a dynamic time warping similarity measure.

Recommended citation: V. Mangé, J.-Y. Tourneret, F. Vincent, L. Mirambell, F. Manzoni Vieira, B. Pilastre, "Détection de Comportements Anormaux dans des Trajectoires de Navires avec One-Class SVM et Dynamic Time Warping", in Proc. Colloque GRETSI, Grenoble, France, 2023
Download Paper