Research On D-S Evidence Theory And Its Application In Automobile Fire Warning

Fire warning system has been widely used in the automotive industry,and plays a very important role in protecting automotive safety.However,the traditional fire warning system uses a single sensor to collect a single fire characteristic quantity,and according to the set threshold to judge.It is easy to cause misjudgments and missed judgments.How to improve the accuracy of automobile fire decision-making and reduce the loss of fire is the key of this research.At present,the main problems of the fire warning system are single detection mode,high false alarm rate and misjudgement rate,on the basis of collecting several automobile fire characteristic parameters,in order to improve the reliability and accuracy of automobile fire warning,this paper applies multi-sensor information fusion technology to fire warning system,by utilizing the advantage of uncertain and incomplete information in multi-sensor information fusion technology processing system.The main contents of this paper are as follows:Firstly,by consulting a large number of relevant literature,the research status of automobile fire early warning system and multi-sensor information fusion technology at home and abroad are analyzed and summarized,and the causes and characteristics of automobile fire are also studied,Based on the analysis of multi-sensor information fusion algorithm,the algorithm-D-S evidence theory in this paper is established,and the basic concepts and combination rules of D-S evidence theory are mastered.Secondly,the reasons of the paradox in D-S evidence theory are studied in depth,and also summarizes different scholars to measure conflict evidences from different perspectives.A modified D-S evidence theory method based on angle cosine and Random distance is designed.The effectiveness of the improved algorithm is verifiedby numerical simulation experiments.Thirdly,the hierarchy of multi-sensor information fusion is introduced,and the decision-level information fusion model of D-S evidence theory is established.By analyzing the characteristic parameters related to the occurrence of automobile fire,a basic probability assignment function based on fuzzy membership degree is constructed for feature level fusion.The membership parameters are determined by combining expert experiences and weighted average method.Finally,the membership functions of each characteristic parameter are obtained.Finally,the overall framework of automobile fire warning system is built and the experimental environment is analyzed.In order to get the original data,the characteristic parameters of automobile fire under different modes at the same time are collected.Then,by constructing the basic probability assignment function,the improved D-S evidence theory is used to make decision on the results under different modes and verify its reliability.Experiments show that the improved method can deal with highly conflicting evidences in the system,and can be well applied to the automobile fire warning system.