The Labeled Multi-Bernoulli Filter for Jump Markov Systems Under Glint Noise

Abstract

This paper proposes a novel labeled multi-Bernoulli (LMB) filter for jump Markov systems (JMS) to track the multiple maneuvering objects under glint noise. By modeling the glint noise as a Student's t-distribution and using the variational Bayesian method to acquire the approximate state distribution, we present an efficient implementation of the LMB filter with joint prediction and update for JMS. Simulation results illustrate that the proposed filter outperforms the existing filters for multi-object tracking under glint noise.INDEX TERMS Multi-object tracking, glint noise, labeled multi-Bernoulli filter, variational Bayesian method, jump Markov systems.

Marginal Distribution Multi-Target Bayes Filter With Assignment of Measurements

“…We exploit the linearization strategy of nonlinear function to deal with the nonlinearity of observation equation 47, and replace the predicted observation vector H k m i,k|k−1 and observation matrix H k in (33), (34), (35) and (36) with the predicted observation vector h(m i,k|k−1 ) and Jacobian matrix H i,k respectively, where…”

“…One is the labeled multi-Bernoulli (LMB) filter [28] and the other is the efficient implementation of the generalized labeled multi-Bernoulli (GLMB) filter [30] called the rapid GLMB (R-GLMB) filter in this paper. The extensions of the δ-GLMB filter for diverse applications have also been reported in [31]- [34]. Despite its efficiency, Vo's efficient implementation of the δ-GLMB filter [30] still requires a much larger computational load than the PHD filter or CBMeMber filter due to the fact that the number of hypothesized targets or tracks in this implementation is significantly greater than the number of real targets.…”

Adaptive δ-Generalized Labeled Multi-Bernoulli Filter for Multi-Object Detection and Tracking

“…Recently, the δ-generalized labeled multi-Bernoulli (δ-GLMB) filter was proposed in [10] and [11]. Its efficient implementation called the rapid GLMB (R-GLMB) filter [12] and efficient approximation called the labeled multi-Bernoulli (LMB) filter [13] were developed to reduce the high computational complexity of the δ-GLMB filter.A number of extensions of the δ-GLMB filter for diverse appli-cations have been reported in [14]- [21] to track the space debris [16], spawning object [17], multiple maneuvering targets [18], targets under glint noise [19], extended targets or group targets [20], and multiple weak targets [21]. The main advantages of the R-GLMB filter over the PHD filter and CBMeMBer filter are that it may provide object tracks and that it is applicable to the case of high clutter density and low detection probability [12]- [13], [22].…”

Multidimensional Information Fusion in Active Sonar via the Generalized Labeled Multi-Bernoulli Filter

“…RFS is rigorous, elegant, and suitable for the multidimensional heterogeneous information fusion and semantic target information fusion scenarios to which traditional methods are difficult to apply. Led by Mahler and Vo-Vo, a group of outstanding scholars engaged in this work and successively proposed the implementation methods of probability hypothesis density (PHD) filter [11] and its cardinalized version, cardinalized PHD (CPHD) [12], the multi-target multi-Bernoulli (MeMBer) filter [8], [13], and the newly derived generalized labeled multi-Bernoulli (GLMB) filter [14]- [16] and its special case, the labeled multi-Bernoulli (LMB) filter [17]- [19]. Among them, GLMB filtering is based on the labeled RFS theory and strictly derived, which has better cardinality estimation accuracy and OSPA metric [20], [21] performances than PHD, CPHD, and MeMBer.…”