基于聚类的多目标自适应互联跟踪算法

Abstract

In data fusion systems, multitarget data association is an important and difficult field of research. Currently, joint probabilistic data association (JPDA) is a common approach to solve this problem; however, the confirmed matrix is generated according to all measurements available at each time step. Therefore, in real situations involving dense clutter and multiple targets, it will result in increased computational load and unacceptably poor performance. To address this problem, an adaptive algorithm for multitarget tracking (adaptive-JPDA) is presented based on clustering. First, a clustering analysis step is applied to divide the measurements into different clusters. Second, the association approaches are selected adaptively according to the measurement parameter; in this way, each measurement can be associated with the correct track or clutter. In order to track multiple targets, the joint association matrix is constructed according to the membership matrix and the relationship matrix, which results in joint association probability. Third, a theoretical analysis is presented to compare the computational complexity. The main idea is that without the pre-tracking or matrix splitting process at each time step, all measurements are handled via a clustering step before data association. Thus, the extent of data association can be reduced dramatically, which reduces the computational complexity. Without disturbance from independent measurements outside the clustering, this approach can yield significant improvements in tracking performance. Finally, simulation results demonstrate the effectiveness of this approach.