Multi-Sensor Data Fusion Based on Improved Analytic Hierarchy Process

As an important method for uncertainty modeling, Dempster-Shafer (DS) evidence theory has been widely applied in practical applications. However, the counter-intuitive results are often generated when fusing different sources of highly conflicting evidence with Dempster's combination rule. Several different methods for measuring the evidence conflict have been proposed. Nevertheless, these methods showed focus only on a single criterion to measure the conflicting evidence. Mono-criteria factor for the measurement of the conflict between evidence is, however, often unreliable and inaccuracy. Because various factors affect the degree of conflict between the evidence, such as imperfection, dissimilarity, disparity, and uncertainty. To address this issue, multiple criteria factors are utilized to measure the degree of conflict between the evidence in this paper. An improved analytic hierarchy process is proposed to determine the weights of each body of evidence by considering multiple criteria. Firstly, calculating the quantitative value of the evaluation index of each evidence under every criterion. The covariance matrix of the criterion layer is determined based on the covariance between the quantitative values of each criterion. Then, the pairwise comparison matrix of the criterion layer can be obtained by transforming the covariance matrix. Next, the variance among the quantitative values of each criterion is applied to construct the fuzzy preference relation matrix. The fuzzy preference relation matrix is used to replace the pairwise comparison matrix of the scheme layer. After that, the weight of the criterion layer and the scheme layer are combined to acquire the final weights of each evidence. Finally, the original evidence is modified with the final weights of the evidence before using Dempster's combination rule. Two numerical experiments are given to verify the efficiency of the proposed approach. The result shows that the proposed method is more efficient and feasible in managing the conflicting evidence than other approaches available in the literature described.