Abstract
This paper reports an adaptation of the GM-PHD filter for ground moving target tracking. In particular, a technique for modeling the measurement covariance matrix adaptively based on the signal power of each detection is investigated. This technique is compared to the standard method of using a fixed covariance matrix both by simulation and experimental data. Notably, the analysis of the error distribution within the experimental data set shows the improvement of the measurement model due to the introduced technique. Simulation and results with experimental data also show that tracking with adaptive measurement covariance matrices yields results that are at least equal to tracking with the best choice of a fixed measurement covariance matrix. Finally, figures with tracks of ground truth targets and targets of opportunity are presented.