Research On The Probability-likelihood Product Method Of Bayesian Filtering

Benefit from the development of the supporting discipline theory such as Control Theory,Signal Processing,Artificial Intelligence,Mathematical Statistics,Bayesian filtering algorithm which based on the framework of Bayesian estimation made a great progress in decades,and has been applied extensive and profound in modern military and civil fields.The development of Bayesian filtering,promote its application in the practical problem,on the other hand,in order to meet or have appeared in the application of various problems may occur,and need to further improve the research of Bayesian filtering,The research studies various Bayesis filtering algorithms under the background of target tracking,which mainly concentrated in the following contents:Firstly,the problems of various filtering algorithm under the framework of bayesian theory in linear and nonlinear dynamic filtering are studied,like low accuracy,high computational complexity and the limitation of the application scenario.in view of the UKF filter with low precision,PF filter with high precision but disadvantages of computation and not easy to applied to the lack of a priori information of multi-sensor fusion and UIF problems can be well applied to multi-sensor fusion but also has the problem of low accuracy,this paper proposes a new Bayesian filtering method,probability product realization method,the method has high accuracy,simple calculation,strong real-time performance,easy to engineering application and also suitable for the lack of a priori information of multi-sensor fusion problems,etc.by the comparison experiment with traditional methods and new methods in different motion model and the realization of dynamic system,verify the effectiveness of the new method.Furthermore,the shortcomings of the new method and the improvement of the new method are discussed.Finally,the application prospect of the new method is prospected.