A method for peak extraction in PHD particle filter

In Probability Hypothesis Density (PHD) particle filter for tracking multiple targets, a new peak extraction method was studied. Firstly, we draw the largest weight particle, when the sum of all particle weights in its neighborhood is larger than target weight, the particle is considered as a peak. Then, we do the same steps for other particles, till we find all N peaks. Experiments show that in the same conditions, the new algorithm can extract peaks more correctly than Expectation Maximization (EM), especially for contiguous targets.