Unscented particle implementation of cardinality balanced multi-target multi-Bernoulli filter
Hao Qiu, Gaoming Huang, Jun Gao · 2014
The cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter is an effective multi-target tracking algorithm proposed recently. This contribution applies the unscented particle framework to the implementation of CBMeMBer filter. Importance sampling density function is extended to a higher dimensional space instead of the original Markov state transition density. By taking the latest measurements into account, the UPF-CBMeMBer is able to improve the particle degradation problem and estimation accuracy of target state. It can be seen from the experiments that the UPF-CBMeMBer filter outperforms the original particle implementation of CBMeMBer.