Numerical comparison of objects’ tracks initiation algorithms

In many problems of the initialization of objects’ tracks, changepoint detection algorithms can be used. In the past when computational complexity was an issue, the K/N algorithm gained its popularity due to computational simplicity. A substantially moreefficient track initiation algorithm can be built based on the sequential change detection technique. In this paper, we consider the Finite Moving Average algorithm. We compare the performance of the K/N algorithm with the Finite Moving Average algorithm. The optimality criterion is to maximize a probability of correct detection in a certain time interval under the given false alarm risk measured as a local probability of a false alarm. For performance, we obtain a theoretical estimate and an estimate by Monte Carlo simulations. The results show that the Finite Moving Average algorithm performs significantly better than the K/N procedure.