Research On Multi-sensor Information Fusion Algorithms Based On D-S Evidence Theory And SVM

In recent years,with the continuous development of electronic information technology,the form of information is complex and diverse,the content is dazzling and the quantity is vast.A single sensor to collect information with one sidedness,multi-sensor information fusion technology came into being,the world's investment in human resources to carry out research,more extensive application prospects.This paper mainly focuses on the research of multi-sensor information fusion algorithms of D-S evidence theory and SVM,and makes corresponding improvements to the algorithms.This paper makes a detailed analysis on the multi-sensor information fusion technology,and the fusion principle,fusion category,fusion characteristics,fusion level and fusion structure are discussed respectively.On this basis,this paper studies and improves the multi-sensor information fusion algorithms of D-S evidence theory and SVM..First,this paper analyzes the reason that the decision-level of classical D-S evidence theory algorithm used in the synthesis of high conflict evidence will produce inconsistent with the facts.To solve this problem,on the premise of affirming the correctness of the combination rules,a new evidence measurement standard(Pcor)based on Pignistic probability function and correlation coefficient is proposed.This standard is compared with the classical weighing standard,which solves the shortcomings of the classical weighing standard in some cases.The simulation results show that the new evidence measurement standard(Pcor)not only can effectively measure the conflict of evidence,but also has strong convergence.Secondly,a weighted evidence combination method based on Pcor is proposed.In this method,Pcor is used as a measure of conflict evidence to determine new evidence trust,and the use new evidence trust to construct trust support matrix.A new weight coefficient of evidence is defined by support matrix,and according to the evidence of evidence to modify the weight coefficient.Finally,D-S combination rules are used for combination.The improved algorithm effectively solves the defect of D-S evidence theory.The effectiveness of the improved algorithm is verified by simulation examples.Finally,based on forest fire data,this paper studies SVM algorithm.Firstly,forest fire data collected by various sensors are introduced and analysed.According to the analysis,the SVM algorithm is selected recognition algorithm for forest fire.Secondly,according to FCM algorithm,the training set which can contain the characteristics of forest fire data and has randomness is selected.Then the SVM algorithm is applied to forest fire recognition,and that the algorithm has a high recognition rate by experiment,but the choice of appropriate parameters is very difficult,affecting the performance of fusion and fusion precision.According to this defect,the author proposes a forest fire recognition algorithm which combines improved cuckoo search algorithm(ICS)and SVM(ICS-SVM).This method makes full use of the ability of ICS algorithm to find the best solution and find the most accurate classification parameters of SVM algorithm.The simulation results show that the improved algorithm not only improves the accuracy of parameters,but also greatly improves the recognition accuracy compared with the SVM algorithm.