Histogram-PMHT for fluctuating target models

摘要

The histogram-probabilistic multi-hypothesis tracker (H-PMHT) is an efficient multi-target tracking approach to the track-before-detect problem. A fundamental feature of the H-PMHT is the discretisation of the energy in the sensor data and the assumption of a multinomial measurement model on the resulting image. A problem with the H-PMHT is that the multinomial measurement model fails to account for fluctuations in the target amplitude, which can degrade performance in realistic sensing conditions. The authors propose an alternative measurement model based on a Poisson mixture process to allow for fluctuating target amplitudes. Simulations show that this new approach, referred to as the Poisson H-PMHT, gives more accurate signal-to-noise ratio estimates than the standard H-PMHT, particularly for scenarios featuring targets with fluctuating amplitude.