Generalized optimal sub-pattern assignment metric

In this paper, we present the generalized optimal sub-pattern assignment (GOSPA) metric on the space of sets of targets. This metric is a generalized version of the unnormalized optimal sub-pattern assignment (OSPA) metric. The difference between unnormalized OSPA and GOSPA is that, in the proposed metric, we can choose a range of values for the cardinality mismatch penalty for a given cut-off distance c. We argue that in multiple target tracking, we should select the cardinality mismatch of GOSPA in a specific way, which is different from OSPA. In this case, the metric can be viewed as sum of target localization error and error due to missed and false targets. We also extend the GOSPA metric to the space of random finite sets, and show that both mean GOSPA and root mean squared GOSPA are metrics, which are useful for performance evaluation.