Knowledge-aided multisensor data fusion for maritime surveillance

Multi sensor fusion techniques are widely employed in several surveillance applications (e.g., battlefield monitoring, air traffic control, camp protection, etc). The necessity of tracking the elements of a dynamic system usually requires combining information from heterogeneous data sources in order to overcome the limitations of each sensor. The gathered information might be related to the target kinematics (position, velocity), its physical features (shape, size, composition) or intentions (route plan, friend/foe, engaged sensor modes, etc). The combination of such heterogeneous sensor data proved to benefit from the exploitation of context information, i.e., static and dynamic features of the scenario, represented in a Knowledge Base (KB). A Geographic Information System (GIS) is a typical example for a KB that can be exploited for the enhancement of multi sensor data fusion. The present paper describes potential strategies for "knowledge-based" data fusion in the a rea of Maritime Situational Awareness (MSA). MSA is founded on the data from heterogeneous sources, including radars, Navigation Aids, air- and space-based monitoring services, and recently-conceived passive sensors. Several strategies for optimally fusing two or more of these information data flows have been proposed for MSA applications. Relevant KB information comprises port locations, coastal lines, preferred routes, traffic rules, and potentially a maritime vessel database. We propose mathematical models and techniques to integrate kinematic constraints, e.g., in terms of navigation fields, and different object behaviour into a data fusion approach. For an exemplary sensor suite, we evaluate performance measures in the framework of centralised and decentralised fusion architectures