An Improved MHT Method Based on Kernel Density Clutter Estimation

The traditional multiple hypothesis tracking (MHT) algorithm assumes that the clutter intensity is known a priori. In the observation scene of unknown clutter, the clutter intensity error will lead to a sharp decline in the accuracy of data association. To solve this problem, this paper proposes an improved MHT method with clutter estimation based on kernel density estimation (KDE). Firstly, the kernel density function is used to fit the unknown clutter spatial distribution, and the clutter intensity in the gate at that time is estimated adaptively; then the score function of the track hypothesis is calculated by using the obtained clutter intensity, which improves the accuracy of data association and the stability of target tracking. Simulation results show that MHT-KDE algorithm can effectively improve the track continuity and reduce the number of false tracks in unknown clutter observation scene.