摘要
The original Probability Hypothesis Density (PHD) filter is a tractable algorithm for Multi-Target Tracking (MTT) in Random Finite Set (RFS) frameworks. In this paper,we introduce a novel Evidence PHD (E-PHD) filter which combines the Dempster-Shafer (DS) evidence theory. The proposed filter can deal with the uncertain information,thus it forms target track. We mainly discusses the E-PHD filter under the condition of linear Gaussian. Research shows that the E-PHD filter has an analytic form of Evidence Gaussian Mixture PHD (E-GMPHD). The final experiment shows that the proposed E-GMPHD filter can derive the target identity,state,and number effectively.
The original Probability Hypothesis Density （PHD） filter is a tractable algorithm for Multi-Target Tracking （MTT） in Random Finite Set （RFS） frameworks. In this paper, we introduce a novel Evidence PHD （E-PHD） filter which combines the Dempster-Shafer （DS） evidence theory. The proposed filter can deal with the uncertain information, thus it forms target track. We mainly discusses the E-PHD filter under the condition of linear Gaussian. Research shows that the E-PHD filter has an analytic form of Evidence Gaussian Mixture PHD （E-GMPHD）. The final experiment shows that the proposed E-GMPHD filter can derive the target identity, state, and number effectively.
基金
Supports in part by the NSFC (No. 60772006, 60874105)
the ZJNSF(Y1080422, R106745)
NCET (08- 0345)
