N-scan δ-generalized labeled multi-bernoulli-based approach for multi-target tracking

Abstract

The δ-GLMB based filter has been proposed as an analytical solution to Bayesian multi-target trackers. The δ-GLMB filter has various weighted GLMB components in order to estimate target states. This filter performs pruning according to each GLMB component weight. However, with respect to different uncertainties for example noisy measurements, the weight of GLMB component may decreases and the track of that GLMB is lost in some steps. In this study, the author benefits from N last history of the GLMBs weight to enhance the performance of δ-GLMB filter in more uncertain conditions. To study the efficiency of the proposed method it is applied on a simulation scenario. The experimental results shows improvements in more uncertain conditions.