Online Learning Method Of Maneuvering Target Tracking Filter

Based on the basic Kalman filter algorithm,this paper mainly studies the on-line learning method of maneuvering target tracking filter,and extends it from single maneuvering target to multi maneuvering target tracking filter and verifies it.First of all,according to the analysis of the advantages and disadvantages of the common maneuvering target tracking filter algorithm,an online optimization feedback maneuvering target tracking filter algorithm based on Elman neural network is proposed.Using Elman neural network on-line learning method,the size of the target maneuver is monitored in real time by online learning of the residuals of the target state prediction and the optimal estimation,innovation and filtering gain matrix,and the optimal estimation and motion model parameters are adjusted in real time according to it.The proposed method overcomes the problem of filter divergence that may occur in general unscented Kalman filtering,and compared with the IMM_ELM online learning tracking filtering method,the tracking filtering precision has a more obvious improvement.Secondly,aiming at the problem of lagging optimization in the optimization filtering methods of different maneuvering targets,a tight coupling tracking filter algorithm based on Elman neural network is proposed,which advances the optimization process to the prediction stage in the filtering process,The phase has a relatively optimized prediction value and adjusted noise variance,thus reducing the error caused by the lag of the optimization process.Compared with other traditional optimization methods,the filtering tracking accuracy is improved to some extent.Finally,for the multi maneuvering target association tracking and filtering process in a clutter environment,the association measurement after data association still contains the problem of relative maneuvering error.An online multi maneuvering target tracking algorithm based on the tightly coupled model is proposed.Tightly coupled algorithm model is introduced into multi maneuvering target tracking system,which makes the state prediction value can be timely optimized and corrected,thus reducing the filtering error.Experiments show that the accuracy of the tracking filter after introducing the tightly coupled model is better than that of the general data association filtering algorithm.