On the Quality of Optimal Assignment for Data Association

Abstract

In this paper, we present a method based on belief functions to evaluate the quality of the optimal assignment solution of a classical association problem encountered in multiple target tracking applications. The purpose of this work is not to provide a new algorithm for solving the assignment problem, but a solution to estimate the quality of the individual associations (pairings) given in the optimal assignment solution. To the knowledge of authors, this problem has not been addressed so far in the literature and its solution may have practical aspects for improving the performances of multisensor-multitarget tracking systems.

The Impact of the Quality Assessment of Optimal Assignment for Data Association in a Multitarget Tracking Context

“…Among the well known are KuhnMunkres algorithm (known as Hungarian) [12,13] and its extension proposed by Bourgeois and Lassalle in [14] to rectangular matrices. More sophisticated Murty's method [15] provides not only the first best assignment, but also the m-best assignments in order of increasing the cost, as shown in the examples of [10,11].…”

“…In such a case, it will be more cautious not to rely on all the pairings confirmed in the first best solution, but only on some of them which are trustable enough, according to the apriori defined threshold level. Utilizing the already obtained and available m-best assignments solutions, Dezert et al [10,11] provide a very efficient method for achieving this important knowledge.…”

“…The method [10] recommends to use the Proportional Conflict Redistribution rule No 6 (PCR6), proposed originally in DSmT framework [8], because it has been proved very efficient in practice. With PCR6, the following fusion result…”

“…The most recent method proposed by Dezert & Benameur [10] to evaluate the Quality Assessment of Data Association (QADA) encountered in multiple target tracking applications in a mono-criterion context, and recently extended in [11] for the multi-criteria context, deal just with the case above given. It assumes that the rewards matrix is known and has been obtained by a method chosen by the user.…”

“…Section 2 describes the problem of DA in multitarget tracking context. Section 3 provides details about the new method [10] for quality assessment of an optimal DA solution. In Section 4 the simulation scenario and results are presented and discussed.…”