Optimization Of Interacting Multiple Model Algorithm Based On Hidden Markov Model

Target tracking is one of the branches in the information fusion field, which has been widely used in military and civil areas. As Aircraft’s maneuvering performance is more and more strong, in order to complete the goal of effective tracking, the importance of the algorithm is highlighted and research in the field of target tracking algorithm is of great significance. Interacting Multiple Model (IMM) algorithm is the most effective method for the study of complex targets tracking’s optimal mixed estimation problem, but there still exists problems such as filtering divergence for the strong maneuvering target estimate. Therefore, this paper puts forward a PTHMM (Probability Transition Hidden Markov Model) method based on maneuvering detection index, who calculates the Markov transition probability matrix parameters dynamically to achieve the optimization and improvement of IMM algorithm under the strong maneuvering state, so as to improve the target tracking performance.(1) For IMM model selection problem in target tracking algorithm, we do a contrastive research of maneuvering and non-maneuvering target models, and compare tracking performance and applicable scenario of the CV, CA and Singer models under target tracking motion mode. The importance of model selection for target motion tracking performance is verified by the simulation.(2) Related factors is analyzed which affects the tracking performance in IMM algorithm, namely the Markov chain state transition probability matrix and IMM initial probability coefficient, etc. Through the simulation, it shows the advantages and limitations of IMM algorithm on tracking performance.(3) As IMM algorithm exists the filtering divergence problem in strong maneuvering target tracking with limited model sets, a PTHMM method is proposed based on maneuvering detection index, where chi-square detection is used as observable state and Markov transition probability matrix as hidden state to build a HMM (Hidden Markov Model, HMM) model. Transition probability matrix’s weights are calculated by Viterbi algorithm. The proposed method modifies the original transition probability matrix of IMM algorithm dynamically to realize the optimization of IMM. The simulation result shows that the improved IMM algorithm reduce the computational complexity, improve the strong maneuvering target tracking’s accuracy and model switching speed.