Spatial Clutter Measurement Density Estimation with the Clutter Probability for Improving Multi-Target Tracking Performance in Cluttered Environments

Abstract

The measurements obtained from a target tracking sensor include the clutter measurements as well as measurements of the targets being tracked. In order to perform target tracking in this environment, a data association method that can effectively distinguish the target measurement from the clutter measurements is required. Target tracking algorithms with data association are designed based on the assumption that the spatial probability distribution of the clutter measurements is uniform in general. However, in real environments, since the prior information about the clutter measurement is unknown, target tracking may not perform properly if this assumption is violated. Therefore, the spatial clutter measurement density estimator (SCMDE) for estimating the clutter measurement density as a non-parametric variable has been proposed. As the SCMDE estimates the sparsity by considering that the adjacent measurements are generated from clutter, the sparsity may be biased if the target measurement is located close to the measurement of interest. Moreover, target tracking performance may be degraded in the situation where multiple targets intersect. In this paper, we propose a clutter measurement density estimation method that can more accurately estimate the clutter measurement density by calculating the clutter probabilities of the adjacent measurements. The proposed method is applied to multiple target tracking to verify the performance improvement by a series of Monte Carlo simulation.