Object correlation evaluation for location data fusion

In a complex tracking environment like an airport many objects are tracked simultaneously. Some of these objects correlate to each other naturally, and thus the questions are how this correlation can be detected and how it can be used for increasing accuracy of the localisation system. The detection of the correlation is done by deciding if two objects are physically or logically bound or connected with each other. This could for example imply that a person is carrying a suitcase or sitting in a car, if these objects are tracked in a localisation system. The proposed approach to answer this question follows a two-step approach by checking computational-inexpensive preconditions first and evaluating a test statistic afterwards. This test statistic must be independent on the used fusion filter like Kalman or Particle Filter. Our results show, that one can reliably detect the physical binding and unbinding events and furthermore increase the location accuracy.