Impact of measurement-to-track data association errors on RCS-based target classification

This paper develops a method for quantifying the impact of measurement-to-track data association errors on the performance of RCS-based classification schemes. First, it develops a means for assessing the impact of misassociations on the estimated mean and variance of the track SNR. This is done for both single-target and multiple-target tracking scenarios. Then, it develops an approach for quantifying the impact of errors in the estimated mean and variance on classification performance. As such, the user can easily relate the number of misassociations to the probability of error in the classification scheme, given specific target information and a particular decision region. Furthermore, this approach could prove useful to those designing RCS-based classification schemes. For example, the decision region could be intelligently modified to reduce the overall probability of error if certain types of misassociations are anticipated. Future work will explore revisions necessary to accommodate multi-modal SNR distributions and more sophisticated SNR estimators.