Research On The Technology For High Maneuvering Target Tracking

This dissertation researches on the theory and algorithms for high maneuvering target tracking.The background is introduced and the previous research in maneuvering target tracking is reviewed briefly. The problems in maneuvering target tracking are discussed, and the research content of this dissertation is introduced finally.The algorithm of high maneuvering target tracking based on single model is studied in chapter 2. Firstly, the problems in the current state model and Jerk model are discussed, and the tracking performance is correlated with model parameters (maneuvering frequency and Jerk frequency).The filtering model of parameter is set up by making use of the measurement information. After derives a method for adaptively estimate the parameters, the improved current state model andη-Jerk model are proposed. The new models can be united with UKF easily and derive two new algorithms for high maneuvering target tracking.The state transition matrix and state error variance are correlated with maneuvering frequency and driving white noise. Without setting up parameter model, the algorithm for maneuvering frequency and driving white noise estimation is proposed by making use of the AR filter and MA filter. The new estimate algorithm can be united with Singer model easily and derives MS model for high maneuvering target tracking.The algorithm of high maneuvering target tracking based on multiple models is studied in chapter 4. To decrease the man-made effects on target tracking accuracy, a new algorithm to estimate the transition probabilities of Markov model is presented, and the interacting multiple models tracking algorithm based on time-varying Markov state transition probabilities is designed. A practical method for model set selection is presented by simplifying the conditions which proposed by X.R.Li, and an interacting multiple model(IMM) tracking algorithm based on time-varying Markov state transition probabilities which adopts the adaptive model set is proposed.The theory of high maneuvering target tracking based on Bayesian estimation is studied in chapter 5. The keys for the performance of particle filter are analyzed. A new proposal distribution is adopted by combining the Unscented Kalman Filter with the improved adaptive strong tracking filter. A new high maneuvering target tracking algorithm is proposed by combining the new proposal distribution and IMM.Summary of this dissertation is made in chapter 6, the present problems and further research are pointed out.