Research On Maneuvering Target Tracking Algorithm Based On Smoothing Filtering

As one of the important research directions in the field of target tracking,maneuvering target tracking technology has received wide attention in both domestic and international academic circles,and is closely related to the aerospace field,military field,transportation field,etc.It has high application value.In this paper,based on the interactive multi-model algorithm and Gaussian process regression,the research on the filtering and smoothing methods for maneuvering target tracking is carried out on this basis,and the main work is as follows.1.An interactive multi-model algorithm based on KL divergence is proposed for the problem of scattered estimation caused by the mutual correlation between the models of the traditional interactive multi-model algorithm.The KL divergence theory is applied to data fusion,and the inter-correlation problem between models is solved by replacing the fusion in the interactive multi-model algorithm by weighted KL divergence fusion.And for the problem that the tracking accuracy of the interactive multi-model algorithm decreases in the complex maneuvering field,an interactive multi-model based smoothing filtering algorithm is proposed to improve the tracking effect of the interactive multi-model algorithm in the complex maneuvering scene by using more measurement information through fixed interval smoothing,while using the approximation method to avoid the need to solve the inverse matrix of the measurement noise covariance in the smoothing process,which enhances the stability of the proposed algorithm is enhanced.The experimental results show that the proposed algorithm can effectively track maneuvering targets in complex maneuvering scenarios and has good robustness and accuracy.2.The proposed algorithm is based on the Gaussian process regression,which can be used to track the target when the system process noise is unknown and abnormal measurements occur.The algorithm can track the maneuvering target when the system process noise is unknown and can handle anomalous measurements at the same time.The algorithm sets the conjugate prior of the process noise covariance,introduces the hidden variables,and uses a variational Bayesian framework to iteratively estimate the state,the process noise covariance,and the introduced hidden variables to solve the problem of unknown system process noise.In the iterative process,Gaussian process regression and sliding windows are used to predict the quantiles,and appropriate thresholds are selected according to statistical laws.When abnormal quantiles are found,the predicted quantiles from Gaussian process regression are used to continue the iterative filtering process instead of the abnormal ones to improve the robustness of the algorithm.The experimental results show that the proposed algorithm can well solve the problems of unknown system process noise and abnormal measurements,and maintains good tracking results in complex scenarios.3.The maneuvering target tracking algorithm based on Gaussian process and smooth interaction multi-model is proposed for the problem that the traditional maneuvering target tracking algorithm cannot effectively use the historical trajectory information.The Gaussian process regression is used to predict the time-series data,the predicted trajectory is matched with the historical trajectory,the predicted trajectory and the Gaussian process regression model are updated using by the matched historical trajectory information,and then the updated trajectory is smoothed using by the smoothed interaction multi-model algorithm to obtain the final smoothed trajectory.Through experiments,it is shown that the proposed algorithm can achieve good tracking and smoothing effects in strong noise scenarios by making full use of historical trajectory information.