Evidence Reasoning Method Research Based On Markov Model

As a method of uncertainty reasoning,evidence theory had been widely used in the field of artificial intelligence,detection and judgment,information fusion,etc.because of its simple and flexible reasoning mecharnism.But if there was a greater conflict among evidences when utilizing fusion formula for evidence fusion,the result might conflict with the fact,or even lead to error.This had become the main problem in the practical application of evidence theory.Based on this,considering the high-efficiency dealing with conflict performance of sequential evidences,this paper presented a new method of evidence fusion.In this paper,firstly historical evidences could be amended by using Pignistic distance as similarity measure.And secondly the deterministic state description in the classic Markov chain was extended to the nondeterministic state description.The past evidences were sampled sequentially according to the sliding window whose width is l,A Markov model was established on these past evidences amended,so that a transition probability matrix could be obtained,which was used to compute the evidential representative.Finally,this representative was combined with itself for l-1 times according to Murphy's combination method.Of course,this method was also fit to parallelly fusion in a step.The new method had an obvious advantage through numerous simulation and contrast experiments.That is to say,it efficiently solved the problem of the combination of conflictive evidences,and kept the robustness and sensibility of combinational result.The above method is a quantitative fusion of the pure digital evidence,and the decision of the natural language form is a qualitative one.For the problem that weighted average operator might not make the best decisions result in 2-tuple linguistic decision,according to the above ideas on evidence Markov model,a Markov model was established on 2-tuple linguistic.So a method of making decision based on 2-tuple linguistic Markov model was proposed in this paper.In practical application,since the sequential evidence fusion method based on the Markov chain had high accuracy and robustness,we would apply it to fuse different multi-feature in the aircraft target recognition.So a method of aircraft target recognition was proposed based on Markov evidence fusion after fusion of five different characteristics.