Conflict Evidence Measurement Based on the Weighted Separate Union Kernel Correlation Coefficient

• Main Authors: Guidong Sun, Xin Guan, Xiao Yi, Jing Zhao
• Format: Article
• Language: English
• Published: IEEE 2018-01-01
• Series: IEEE Access
• Subjects: Belief functions; conflict evidence measurement; correlation coefficient; relational matrix; weight; union kernel
• Online Access: https://ieeexplore.ieee.org/document/8372448/

We develop the weighted separate evidence correlation coefficient as the measurement of the conflict evidence for these reasons that most of the existing evidence measurement: 1) cannot separate the consistent and conflict evidences; 2) do not consider the weight of the focal element; and 3) mix the single subset and union subset focal element. In addition, in the evidence theory only the kernel makes sense, so we need not consider all the other elements with zero belief in the frame of discernment and just use the weighted separate union kernel correlation coefficient as such a measurement. This measurement lies in [-1, 1] instead of [0, 1], which can separates the consistent and conflict evidences more clearly. It takes the weight of the focal element evidence into consideration, which can deal with the situation when the importance of the focal element in each evidence is different. Furthermore, it utilizes the defined kernel and a union kernel relational matrix to separate the single subset and the union subset focal element evidences when constructing the conflict evidence measurement. We compare the proposed conflict evidence measurement with the existing methods by some examples and apply it in a multi-sensor fusion process. Through the comparisons, the validity of the proposed measurement is illustrated in detail.