Estimation of Consistent Cross-Covariance Matrices in a Multisensor Data Fusion

The article investigates the problems relevant to the fusion of data provided by sensors used for detection and tracking. One of the most burdensome problems on any data fusion in these fields of application is the search for the appropriate correlation function between the estimation errors of the tracks from the various sensors. The absence of such a function can lead to excessively optimistic cross-covariance matrices between tracks with consequent inconsistencies and missed associations. The article has the aim to provide an effective solution to this problem. The proposed technique resides to build consistent cross-covariance matrices by means of a tracking system which uses as input the simultaneous associations of the tracks of the involved sensors. Furthermore, a fusion equation, that provides an unbiased and optimal global estimate in the sense of the minimum mean square error, is proposed. The method results simple and fast as well as precise. Finally, the robustness of the algorithms is proved through dedicated simulations.