在证据融合推理时,由于系统中出现未知新命题和传感器收集的证据不可靠,使得Dempster组合规则不适用于冲突证据组合。为了有效地处理)中突证据,提出一种基于非封闭识别框架的改进证据推理方法。首先在封闭识别框架内增加一个表示新命题的集合,将封闭识别框架扩展为非封闭识别框架,给出新命题基本概率指派的赋值方法,然后给出非封闭识别框架下焦元的组合规则。在非封闭识别框架下,依据冲突焦元本身基本概率指派以及证据的可信度信息,将局部冲突在涉及冲突的焦元之间分配,最后通过实例结果表明,与封闭识别框架下的Dempster组合规则结果相比,该方法可以有效识别出新命题。非封闭识别框架下的改进证据推理方法,可以通过在非封闭识别框架下,对新命题与已知证据及已知证据之间的两类冲突信息在局部分配,使最终证据推理结果向合理方向聚焦。
When evidences combine and fusion, because the unknown new propositions appear in system and the sensor gathered evidences are uncertain, the D -S combination rules are inapplicable for conflict evidence combination. In order to manage the conflict evidence effectively, a modified conflict evidence reasoning method based on non-closed frame of discernment is proposed. First, an aggregate denoting unknown new proposition is added in closed frame of discernment, then the closed frame of discernment is expanded to non - closed frame of discern- ment, and the evaluate method of new proposition basic probability assignment is given, the focal element combination method in non - closed frame of discernment is given based on the relations between the new proposition and the known evidence. With this as the foundation, the local conflicts are distributed, on the basis of the basic probability assignment and the reliability of evidence among the conflict focal elements. Finally the distribution method of conflicts is verified through with Dempster combination an example. The result shows that this proposed approach, compared rules in non - closed frame of discernment, is effective in identifying with the result of the new propositions. By the modified evidence reasoning method and closed frame of discernment, two kinds of conflicts of new proposition between known evidences and between known evidences can be distributed locally by non -closed frame. By managing the two reasons of conflict evidence in a comprehensive way, the final evidence reasoning result focus on the reasonable direction. Key