Research Of Multi-target Tracking Algorithm Under Glint Noise

The target tracking has been researched in the past decades due to its wide application in military and civil fields.The existing method of the target tracking usually assumes that the measurement noise is the Gaussian noise,therefore is inapplicable to the case where the measurement noise is the glint noise.The probability hypothesis density(PHD)filter under glint noise is efficient for multi-target tracking.However,this filter is inapplicable to the case of low detection probability.In order to track the multiple targets under glint noise more efficiently,we propose a multi-target Bayesian filter under glint noise in this thesis.Applying the method of multiple models to the proposed filter,we also propose a multi-target Bayesian filter with jump Markov system(JMS)models under glint noise.The main content of this thesis is summarized as follows:1)we introduce the multi-target Bayesian filter based on the FISST,PHD filter and sequential multi-target Bayesian filter(SMB),analyze the characteristics of glint noise in radar target tracking,and discuss the method for using the t-distribution to model the non-Gaussian distribution and glint noise with a long tailed characteristic.2)we propose a multi-target Bayesian filter under glint noise.The proposed filter uses the Student's t-distribution to model the glint noise,use the variational Bayesian method to obtain the approximate distribution of the target state,and uses the Gaussian-Gamma mixture terms to represent the prediction distribution and update distribution of the target state.Based on the proposed filter,we develop the tracking algorithms of multi-target tracking for linear system and non-linear system,respectively.Simulation results demonstrate that the proposed filter is better than the PHD filter under glint noise at tracking multiple targets in case of low detection probability.3)in order to track multiple maneuvering targets by using the measurement with glint nose,we propose the multi-target Bayesian filter with the JMS models under glint noise by applying the approach of the switch multiple models to the multi-target Bayesian filter under glint noise.Based on the proposed filter,we also develop the tracking algorithms for linear system and non-linear system,respectively.Simulation results demonstrate that the proposed filter tracks multiple maneuvering targets more efficiently than the JMS-PHD filter with glint noise in case of low detection probability.