Abnormal Moving Speed Detection Using Combination of Kernel Density Estimator and DBSCAN for Coastal Surveillance Radars

In this paper, we investigate the problem of detection of abnormal moving speeds for a coastal surveillance radar. A new definition of normal moving speeds which is based on the historical radar data of maritime targets on the monitoring area is introduced. The historical radar data is mined by the cell-based method, unsupervised machine learning to obtain the vessel normal moving speeds in the monitoring area. Then a logic rule is applied to detect the abnormal targets. The proposed method is tested with real data from a coastal surveillance radar. The test results show that the false alarm rate (FAR) is equal zero. It is also shown that this kind of anomaly detection can be integrated into a coastal surveillance radar for the detection of the maritime illegal activities.