Data Fusion Decision And Probability Conversion Under Strong Conflict Of Evidenc

In multi-sensor information fusion systems,the information provided by the sensors is generally incomplete,imprecise,ambiguous and even contradictory to each other.That is to say,it contains a large amount of uncertainty.The Information Fusion Processing Centre needs to handle the uncertain information reasonably to achieve the purpose of target identification and multi-attribute decision making.Evidence theory can effectively reason about uncertain information in information fusion systems,and the reasoning process can ensure the integrity of the information.D-S evidence theory can also effectively express and deal with uncertain information.It is crucial for information fusion and scientific decision-making to effectively quantify the contradictory differences in uncertain information between sensors based on quantitative data and qualitative knowledge.The strength of D-S evidence theory lies in the reasonable quantification of information uncertainty,but evidence appears too redundant in some cases to express uncertain information,which can make it more difficult for us to deal with uncertain information.Therefore,the simplification of uncertain information in complex situations is also of great research importance.To address these issues,this paper investigated the conflicting measure in traditional evidence theory,and further extended the study of conflict measure in complex evidence theory,and proposed a new method of decision probability transformation to reduce the uncertainty of information.Firstly,this paper defined a strong conflict measure based on the Chi-Square distance.The new conflict metric was named Belief Chi-Square distance,which concentrated the uncertain information in the single subset of focal element,taking into account difference in belief functions between the single-subset focal element and the difference in plausibility functions between the single-subset focal element.The new conflict metric could not only directly handle evidence with composite focal elements,but also avoid the problem of information loss.On this basis,a new data fusion algorithm was defined by combining the conflict metrics between BPAs and the information content of the evidence itself,which solved the problem of paradoxes when the traditional evidence combination rule fused strongly conflicting evidence.In addition,the robustness of the new fusion approach was analysed by adding noise to the BPA.The new data fusion algorithm was applied to iris classification and fault diagnosis to demonstrate its the effectiveness and practicality with practical cases.Secondly,this paper extended the Belief Chi-Square distance to the complex evidence theory and proposed a new conflict measure of complex evidence theory.After the complex belief function and the complex plausibility function were obtained by complex arithmetic,the difference between complex mass functions was then measured by the complex belief Chi-Square distance.Its rationality was proved through several example analyses.Furthermore,a medical diagnosis decision algorithm was proposed by the complex belief Chi-Square distance and the practical application value of the new decision algorithm was illustrated by comparing it with other methods.Finally,this paper investigated the conversion of the mass function into decision probabilities.The prospect theory value function was extended to define the value function of a single subset focal element.The single-subset focal element value function was used to measure the importance of the single-subset focal element,and then the allocation coefficients were generated according to the importance.The BPA of the composite focal element was assigned to the single-subset focal element to finally generate the decision probability.Additionally,the two properties that needed to be satisfied for the probability conversion were proved and the good conversion performance of the new method was also demonstrated by example analysis.