Tracking Split Group with δ-Generalized Labeled Multi-Bernoulli Filter

Abstract

As target splitting is not considered in the initial development of δ-generalized labeled multi-Bernoulli (δ-GLMB) filter, the scenarios where the new targets appearing conditioned on the preexisting one are not readily addressed by this filter. In view of this, we model the group target as gamma Gaussian inverse Wishart (GGIW) distribution and derive a δ-GLMB filter based on the group splitting model, in which the target splitting event is investigated. Two simplifications of the approach are presented to improve the computing efficiency, where with splitting detection, we need not to predict the splitting events of all the GGIW components in every iteration. With component combination applied in adaptive birth, a redundant modeling for a newborn target or preexisting target could be avoided. Moreover, a method for labeling performance evaluation of the algorithm is provided. Simulations demonstrate the effectiveness of the proposed approach.

Tracking the Splitting and Combination of Group Target With $\delta$ -Generalized Labeled Multi-Bernoulli Filter

“…The probabilities for targets birth are r birth,i = 0.03, i = 1, 2, 3. For the δ-GLMB with adaptive birth, V 0 and initial state x 0 are computed on-line, the expected number of At k th moment, the true extension of the i th group target is X Static model is used for target birth, the performance of the 3 different splitting filters are compared: the group splitting approach based on GGIW-PHD filter (here named GGIW-PHD) used in [50], the static birth model based approach in [51], where ( 39) is taken as the splitting criteria (GG-G0), and the GG-DC-G1 filter, where splitting gate 1 (see (40)) and the prediction of splitting dimension (see (43)) are considered. The optimal sub-pattern assignment (OSPA) error of the state estimation, the estimated target number and the execution time of the 3 filters are shown in Fig.…”

“…The splitting mode referred in [50], [51] is corresponding to the S1 of this paper, here we give the approximation of predicted labeled joint density of split pair derived in [51] (20) where, L ,T,+ denotes the label set split from label .…”

“…( 20), ( 26) or (27), given the multi-target posterior be a δ-GLMB of the form (3), the predicted labeled multi-target density after splitting is also a δ-GLMB. More details could be referenced in [51]. The GGIW-δ-GLMB prior multi-group density could be expressed as eq.…”

“…However, it is not easy to determine the value of U 0 , since we can't know the scale of the group and the distribution density of the targets in the group in advance in a non-cooperative scenario, moreover the distribution density of different groups in a scenario may be inconsistent. Reference [51] has the changing of the ratio of the area to the number of measurements be the splitting criteria, is more flexible. The criteria is given as…”

“…A group target splitting model is proposed in [49], where the state's expected value of the split target is identical to the expected value of the state of the original target, while in [50], the reasonable alternatives about the split target are considered, proposes a more general model. An δ-GLMB filter based implementation of the splitting model proposed in [50] was developed in [51], where the tracks of targets are estimated as well as targets states, the split tracks and the newly born tracks are discriminated by the track labels, kept the continuity of split tracks. The splitting and combination models in [52] considered the scenario where a group split into multi-subgroups (>2) or multiple groups (>2) combined into one larger group at the same time, are modifications of the models proposed in [50], however, only the major axis of the target extent is splittable in this splitting model.…”