A Gaussian Mixture PHD Filter with the Capability of Information Hold

The probability hypothesis density(PHD)filter has been proved to be an efficient method for the multi-target tracking in the presence of false alarms,missed detections and an unknown number of targets.However,in the original PHD filter,a large amount of information of the existing targets will be immediately discarded by the PHD filter once they cannot be detected by a sensor at a given time.To resolve the information loss problem of missed true targets,we modify the predication and update equations of the PHD filter and propose a modified PHD filter with the capability of information hold.A Gaussian mixture implementation of the modified PHD filter for linear Gaussian models is also presented.The simulation results demonstrate that the proposed filter can achieve better tracking performance of multiple targets than the original PHD filter in the presence of missed detections.