Using machine learning to optimize autonomous tracking of vessels by marine radar

Knowledge of vessel activity and human use in sensitive marine areas is essential for effectively managing marine resources. The Marine Monitor (M2) platform integrates X-band marine radar, a high-definition camera, and a meteorological sensor to document vessel activity and environmental conditions in coastal areas, thereby providing a method for continuous and autonomous monitoring. Machine learning was used to classify targets detected by the radar system as true vessels or false targets caused by sea clutter and other confusers. Using a model that incorporated time-averaged weather variables provided by the meteorological sensor, target records were classified with 98% accuracy at two distinct locations. The ability to discriminate between true vessels and false targets can facilitate a more accurate estimate of vessel activity detected by radar.