The sequential PHD filter for nonlinear and Gaussian models

Abstract

The probability hypothesis density (PHD) filter handles the measurements periodically, once a scan period. Since measurements have to be gathered for a scan period before the PHD filter can perform a recursion, significant delay may arise if the scan period is long. To reduce this delay, we proposed sequential PHD filter. A Gaussian mixture implementation of the sequential PHD filter for nonlinear and Gaussian models is also developed, where the unscented transformation is employed to deal with the nonlinearities of target dynamic and measurement models. The simulation results demonstrate that the proposed filter updates the posterior intensity whenever a new measurement becomes available, and tracks multiple targets better than the PHD filter in the presence of missed detections.