Joint detection, tracking and classification of multiple extended objects based on the JDTC-GIW-MeMBer filter

摘要

In this paper, the problem of joint detection, tracking and classification (JDTC) of multiple extended objects (EOs) is addressed using the random finite set (RFS) theory. In the proposed method, the shape of an EO is restricted to an ellipse that can be easily modelled by a random matrix. The classification of an EO is estimated on the basis of size information with respect to its extended part. Furthermore, the extended object RFS is proposed to be modelled as a multi-Bernoulli (MeMBer) process, then the MeMBer filter is naturally adopted to propagate the EO posteriors. In this paper, in order to achieve the closed-form solution, we model the spatial probability density function (SPDF) of each Bernoulli component as the mixture Gaussian inverse Wishart (GIW) process for the JDTC problem. Compared to the existing JDTC strategies, the proposed algorithm, named as JDTC-GIW-MeMBer filter, is demonstrated via simulations to achieve better performance to the existing algorithms on detection and tracking of the EO centers, the estimation accuracy of the extended states, and EO classifications.