Performance measures for multiple target tracking problems

Performance evaluation is one of the most important steps in any target tracking problem. The objective of this paper is to present a brief review of different approaches available for the performance analysis of multiple target tracking algorithms. Metrics are first classified into sensor-dependent and tracker-dependent ones. Then, the tracker-dependent measures are discussed after classifying into two groups named algorithm-free and algorithm dependent measures. For the classification purposes, three different categories of algorithm-free metrics are described. Finally, to demonstrate the application of the metrics in evaluating the performance of different tracking algorithms, a challenging scenario is considered.