Extended Target Fast Labeled Multi-Bernoulli Filter

Abstract

Focusing on the real-time tracking of the extended target labeled multi-Bernoulli (ET-LMB) filter, this paper proposes an extended target fast labeled multi-Bernoulli (ET-FLMB) filter based on beta gamma box particle (BGBP) and Gaussian process (GP), called ET-BGBP-GP-FLMB filter. First, a new ET-FLMB filter is derived to reduce the computational complexity of the ET-LMB filter. Then, by modeling the target state as an augmented state including detection probability, measurement rate, kinematic state and extension state, the BGBP-GP implementation of the ET-FLMB filter is presented. Compared with the traditional sequential Monte Carlo (SMC) implementation, the proposed implementation can not only greatly reduce the number of particles and the amount of computation, but also estimate the detection probabilities, measurement rates and extension states while estimating the number and kinematic states of extended targets. Finally, the simulation results show that the proposed filter can significantly reduce the computational burden and improve the real-time performance.