Evaluation of multistatic tree-search based tracking on the SEABAR dataset

Abstract

The focus of this paper is the extension of tree-search based tracking to multistatic tracking problems and the evaluation of the proposed algorithm on the SEABAR'07 sonar dataset. The tree-search based tracker, originally introduced in, is built upon the stack algorithm for convolutional decoding. To perform track estimation, the tracker navigates a search tree in which each path represents a sequence of states visited by the target. By exploring only a subset of the search tree, the stack-based tracker computes only likely regions of the posterior distribution at each update, thereby approximating the Bayesian inference solution to the tracking problem. In this work, the monostatic stack-based tracker is extended to multistatic tracking. The structure of the tree-search approach facilitates the incorporation of information from multiple source-receiver pairs with minimal complexity increase. The performance of the multistatic stack-based tracker on the SEABAR'07 dataset shows that the tracker is able to maintain track through highly nonlinear target maneuvers and in the presence of heavy clutter.