Automated surveillance track filter tuning by randomized algorithms

Tuning or fine tuning of a tracker system turns out to be a hard job in practice. The main reason for this is that in a practical (surveillance) tracker system there are a lot of design parameters and a lot of competing requirements to be met. This paper provides the user with an algorithm to tune a tracker system automatically and at the same time obtain quantitative results in terms of the optimality of the solution. The theory of randomized algorithms is used to obtain probabilistic statements on the quality of the output of the tuning process. A simplified example illustrates how the developed theory is to be used.