Fishing Gear Identification From Vessel-Monitoring-System-Based Fishing Vessel Trajectories

The surveillance of illegal fishing activities is a critical issue for the management of marine resources. In this study, we investigate the space-based monitoring of fishing vessel activities using vessel monitoring system (VMS) trajectory data. Our specific objective is the automatic recognition of the employed fishing gear type from VMS data. The proposed approach combines the extraction of new VMS-derived features, issued from the nonsupervised identification and characterization of gear-specific movement patterns, and supervised machine learning, namely, random forest and support vector machine. We explore the use of the proposed features jointly to more classical ones (e.g., mean position and sinuosity index). Overall, we reach recognition performance greater than 97% for the considered Indonesian fisheries and present an application to the detection of abnormal fishing vessel behaviors with respect to the registered fishing gear. We further discuss the relevance of the proposed approach and its potential for the operational monitoring of fishing vessel activities.