BPA Generation Based On Projection Method And Conflict Evidence Fusion Decision-making Metho

Multi-sensor information fusion is to analyze and synthesize various types of information to obtain more complex,reliable,and accurate information.D-S evidence theory is one of the most commonly used methods of multi-sensor information fusion,which was first applied in expert systems and has great advantages in dealing with uncertain and unknown information.However,there are still some problems when using evidence theory for data fusion due to external interference or sensor failure.First problem is how to generate the reasonable basic probability assignment(BPA).Second,D-S evidence theory may lead to counterintuitive results when fusing highly conflicting evidence.To address the above problems in D-S combination rules,this paper proposes a new method for generating BPA and two methods for combining conflicting evidence as follows:(1)Based on the theory of bidirectional projection,a new BPA generation method is proposed.First,initial BPAs are generated by the weighted bidirectional projection between the interval number model of the training set and the test set.Then,the initial BPAs are reconstructed based on the distance between the evidences.Finally,the reconstructed BPAs are fused by using Dempster combination rules,and the final result after fusion will be used for multiple-attribute decision-making.Experiments show that the recognition accuracy for Iris dataset based on the proposed method is 96.67%.For other datasets,the experimental results also verify the rationality and reliability of the method.In the case of insufficient samples,the proposed method still has high classification accuracy.(2)A new uncertainty measurement under the discernment of evidence theory is proposed.First,an improved normalized projection method is proposed,and its mathematical properties are proved.Then a new uncertainty measurement method is proposed based on this method,which satisfies all expected properties of uncertainty measurement.Compared with existing uncertainty measurement methods,it avoids the problem of information loss caused by switching between different discernment in traditional methods.Finally,the proposed method is applied to the fusion of conflicting evidence,and numerical examples and practical applications verify that the proposed method can effectively solve conflicting evidence fusion issues.(3)A decision-making method of evidence fusion based on the proof by contradiction and complete frame of discernment is proposed in this paper.Firstly,assuming that each proposition is correct,and make inferences and calculations based on the hypothesis.The distance between each evidence and the corresponding vector of the expected result is calculated,and the support degree of each proposition can be calculated based on the distance.Then the distance formula is used to calculate the distance again between the support degree of the proposition and each evidence.Finally,the final result can be obtained by the weighted method.In comparison with other methods,several experiments verify the effectiveness of the proposed method.