证据的分组合成法

Grouping method for combining evidence

为了提高证据融合的精确度并降低融合的运算量,结合批量式融合和序贯式融合的优点,提出证据分组合成法.该方法首先判断证据间是否可用Dempster组合规则进行合成,若可以,则两证据归为同组;否则归为不同组.对于同组证据,利用Dempster组合规则直接合成,即组内证据序贯式融合;对于不同组证据,通过最优化模型修正各组证据源,再利用Dempster组合规则合成,即组间证据批量式融合.算例分析验证了该方法具有运算量小、稳定性好、精确度高的特点.

To improve the precision of evidence fusion and simplify its computation, a grouping method for combining evidence is proposed to take the advantages of sequence fusion and batch fusion. Firstly, the proposed method judges whether the pieces of evidence can be combined by using Dempster rule of combination. If the pieces of evidence are judged to be fused with Dempster rule of combination, they will be put into the same group. Otherwise, they will be put into different groups. The evidence in the same group is combined directly with Dempster rule of combination, which forms sequence fusion within groups. For the evidence in different groups, an optimization model is developed to modify the evidence, and then combined with Dempster rule of combination, which forms batch fusion among groups. Numerical examples are provided to demonstrate the advantages of the proposed method, such as less computation, better stability and higher precision.