Performance of Track-to-Track Association Algorithms Based on Mahalanobis Distance

In multi-sensor tracking system, the track-to-track association problem is to determine whether a set of local tracks from different sensor systems are represent the same target. This problem is usually formulated as a binary hypothesis test, and the most common statistics is defined as the squared Mahalanobis distance (SMD) between the kinematic state estimates of two tracks. In this paper, three types of SMD algorithms are investigated, i.e., the SMD algorithm, the cumulative SMD algorithm, and the Discrete Wavelet Transform (DWT) algorithm which can be regarded as a generalized SMD ratio algorithm. The first one can be looked as singlescan algorithm, and the rest two are multiscan approaches. From another viewpoint, the first two are time domain algorithms, and the last one is a transform domain algorithm. The probability distribution functions of statistics defined by these algorithms have been discussed under the assumption that the estimates errors are independent across time. The Operating Characteristic Function is used to describe association performance. It shows that the multiscan algorithm performs better than the singlescan algorithm. As to multiscan algorithms, the DWT algorithm is superior to time domain algorithm. But better algorithm is more sensitive to the residual bias because the statistic based on SMD of target state estimates is directly contaminated by the bias. Copyright © 2013 IFSA.