| The probability hypothesis density filter based on random finite set does not have to deal with the problem of data associate and has good tracking performance.So it has become a hot research field in multiple target tracking.The labeled probability hypothesis density filter can estimate not only the states of targets,but also target trajectories,and has high application value.On the basis of the labeled probability hypothesis density filter,this paper mainly researches the multi-target tracking problems under different scenarios.The specific contents are as follows:(1)In order to solve the problem that labeled Gaussian mixture hypothesis probability density filter(LGM-PHD)cannot get the spawn targets,the labeled GM-PHD filter with spawning targets is proposed.Firstly,the labels are applied to the Gaussian items in the GM-PHD filter to distinguish different targets,and their tracks are determined.Secondly,in the process of filtering,the labels of state estimation obtained at each time are matched with the formed track labels to achieve track maintenance.Finally,the spawn threshold is used to determine whether spawn targets exist or not and the number of possible spawn targets.And then the labels for Gaussian items of new targets and possible spawn targets are reallocated.The Simulation results show,in the situation that exist spawn targets,the improved algorithm has better tracking performance than the LGM-PHD filter.(2)In multi-target tracking system,the false measurements increase with the increase of the clutter,which causes the time complexity of the LGM-PHD filter increase.In addition,the estimation performance of LGM-PHD is poorer with the stronger clutter.Aiming at this problem,an improved LGM-PHD algorithm in strong clutter environment is proposed.Firstly,the predicted information and measurement data are used to calculate the residual vectors after prediction.The ellipsoid threshold technology is used to get the effective measurements that closing to the true target state.Secondly,the effective measurements are used to update the Gaussian components.After getting the targets’ states,the label management is used to update the targets’ trajectories and manage the track of target.The simulation results show that the proposed algorithm not only decreases the algorithm complexity but also has better tracking performance.(3)In multi-target tracking system,some targets will be lost when the targets are crossing or closed to each other.In order to solve this problem,an improved LGM-PHD filter with crossing targets is put forward.Firstly,after update step,determining whether the target is lost by managing the estimated targets labels.Then when the target is missed,determining whether the targets are crossing or closed to each other.If their trajectory is closed to each other,managing labels and resetting the weights of Gaussian items,and re-estimating target states and tracks.The simulation results show that the improved algorithm can solve the missed detection caused by the target crossing successfully,and has good stability. |
