Ground Target Tracking Methods Based On Multi-bernoulli Filter

The ground targets have the characteristics of complex background information,strong maneuverability and high clutter density,which causes its state and number to change constantly.The traditional multi-target tracking algorithms take the target association as the core and cannot accurately correlate the ground targets,thus invalidating;The finite set multi-target tracking algorithm expresses the ground targets in a collective form,avoiding the target association and thus adapting to the ground targets.In this paper,the potential-balanced multi-objective multi-Bernoulli filter(CBMeMBer)based on one of the stochastic finite set filtering methods is taken as the main research object,and the research is carried out from two aspects: tracking accuracy improvement and engineering implementation.In the aspect of improving the filtering accuracy,firstly,a variety of nonlinear Kalman filtering methods and the improvement of the filtering effect after combining with CBMeMBer under Gaussian mixture are analyzed and compared.Secondly,for the strong maneuverability of the target,the a priori information is utilized.A statistical model based on the statistical distance between the filtered results is used to design a CKFCBMeMBer based on multi-model mixing,and the elliptic threshold is used to determine the correlation threshold between the targets,so as to improve the overall filtering accuracy of the filter.In terms of engineering implementation,the root mean square volume Kalman filter is combined with CBMeMBer to ensure the continuous filtering process to improve its stability.And the Euclidean distance between the targets and the rectangular threshold are used for targets in different moments.The target establishes a label relationship and removes clutter to achieve the purpose of reducing the computational complexity of the filter.Experiments were carried out on the improved methods of tracking accuracy and engineering implementation.Firstly,the improvement of the overall tracking performance under the condition of combining cubature Kalman filter and CBMeMBer is verified by experiments.By comparing the filtering error of multi-model CKF-CBMeMBer and single model SCKF-CBMeMBer,the filtering accuracy is improved.Experiment results show that the CBMeMBer filtering accuracy based on model mixing is 8.75% higher than that of the single model SCKF-CBMeMBer.Then,stability of the combination of root mean square Kalman filter and CBMeMBer is verified by comparing its filtering effects with different clutter density.On this basis,the processing efficiency of the label-based SCKF-CBMeMBer is compared with common CBMeMBer,which has witnessed a 2.6% decrease in operating time.