Sensor Data Fusion for Target Classification and Identification

Toyon Research proposes to develop a fusion module that fuses measurements collected by an arbitrary number of sensors over time to provide target recognition of airborne targets. The fusion module will consist of a Bayesian Network and a statisticaldatabase. The Bayesian Network will update beliefs about what type of target is being measured by taking into consideration a priori information regarding the types of airborne objects in an area and the capabilities of the sensors providing measurements.Additionally, the network will consider the degree to which a particular measurement matches the expected signal of each type of target of interest. The statistical database will provide information regarding the expected performance characteristics of thesensors and the degree to which a particular measurement matches expectations. One of the important advantages of our approach is that information from sensors which provide data at different levels in a target-class hierarchy can be effectively fused. Forinstance, data from a sensor which distinguishes a jet aircraft from a prop aircraft can be fused with data from a sensor which distinguishes an F-15 from an F-16. We propose to develop an initial version of the fusion module and provide a demonstration ofits capabilities on an example problem during Phase I.The successful completion of this research will result in the development of a fusion module which can fuse measurements collected by an arbitrary number of sensors over time to robustly providenon-cooperative target recognition of airborne objects. Non-military applications of this technology include air traffic control, counter-drug operations, and medical diagnostics.