Density Based Clustering Data Association Procedure for Real–Time HFSWRs Tracking at OTH Distances

In order to efficiently cover maritime areas at over the horizon (OTH) distances and thus increase marine safety in a nation's exclusive economic zone (EEZ), a network of maritime sensors built around High Frequency Surface Wave Radars (HFSWR) can be an excellent choice. The critical parameter for success of the deployed sensor network is a real time tracking of all detected vessels. During the tracking process, data association (DA) is the first step and it defines the complexity and thus the speed of the whole tracking process. This paper presents a density based clustering DA procedure where the cluster complexity determines the applied DA procedure within the cluster itself. It is demonstrated that the great majority of clusters (over 98 % of all clusters in the worst case) may be processed in a timely manner with an optimal DA procedure, or more precisely, a Joint Probability Data Association (JPDA). However, a small number of unusually large clusters (less than 2 % of all clusters) requires the application of a sub-optimal DA procedure, more accurately, the Roecker's suboptimal JPDA algorithm, in order to maintain real time performance of the whole tracking process. Moreover, unlike standard JPDA procedure which tends to be inapplicable for real time tracking in heavily cluttered environment, the density based clustering DA procedure presented here provides real time performances in the very same environment. The whole analysis is done on real HFSWR data obtained from two HFSWR, located in the Gulf of Guinea. The data set used for the experiments includes data obtained during a month and a half of constant HFSWR operation.