D-S Evidence Theory Based On Belief Coulomb Force And Its Application

Dempster-Shafer(D-S)evidence theory has great advantages on dealing uncertain information and been widely applied.However,the evidence could be conflicting or inconsistent in practical applications.The measurement on these conflict evidence has been paid much attention by a large number of researchers.This article has conducted an in-depth discussion on this problem in DS evidence theory,and proposed a concept of Belief Coulomb Force(BCF)as the evaluation on the conflict evidence.The effectiveness of the presented method was proved by the application of panoramic plant classification and identification.Practicability,the main tasks are as follows:Firstly,this paper proposed a DS improvement method based on probability magnification to correct conflict evidence.In detail,the conflict of the evidence would be amplified simultaneously,and then the weight of each piece of evidence is determined according to the Euclidean distance between the individual evidence and the average evidence,and the average weight is used as the threshold to judge conflict evidence,and finally the conflict evidence is amended based on the weight of the conflict evidence.Secondly,a BCF-based D-S evidence theory was proposed.Initally,based on Coulomb's law,the concept of BCF was proposed,and Zhou entropy was used to describe the amount of charge.It took the scale of the recognition frame(FOD)and the relative scale of the focal element relative to the FOD into account to better represent the amount of information of evidence.Then,applying the Jiang conflict coefficient to express the distance between belief functions and using the Pearson coefficient to identify whether the Coulomb force between the evidence is gravitational or repulsive.The proposed BCF aims to analyze the process of multisource information fusion from the perspective of electromagnetics,so as to provide new insights for resolving the conflict management of evidence theory.Finally,a fault diagnosis example was displayed to validate the D-S evidence theory based on the belief Coulomb force.Thirdly,a multi-source information fusion recognition method for panoramic plant images based on the BCF-based D-S evidence theory was proposed.The graylevel histogram(GH),wavelet transform(WT),and gray-level co-occurrence matrix(GLCM)were used to extract the texture features of panoramic plants,and the BP neural network classifier was used for parallel training and testing.The results showed that the recognition accuracy of plant features extracted by a single method would reach about 70%,while the DS evidence theory,the DS evidence theory modified by the amplified-probability conflict evidence,and BCF-based DS evidence theory were used to identify the three characteristics.After fusion,the recognition accuracy rates can attain 80.504%,82.28%,and 85.71% respectively.The experimental results demonstrated that BCF-based D-S evidence theory is superb among the above methods to identify panoramic plant images,and it can effectively reduce the uncertainty information in the recognition process and improve the recognition accuracy.And the practicability of the proposed method can be proved.