基于矩阵分析的一种不确定性推理的数据融合方法

采用不确定性推理的矩阵分析方法,建立了用于目标识别的多传感器数据融合的数学模型. 该模型综合了来自多种不同传感器的基于正态分布的检测数据,通过定义相关系数矩阵来获取基 本可信度分配值矩阵. 提出了一种多传感器信息融合的新算法,该算法依靠可信度的积累,通过多 级递推融合可获得目标状态基于全局信息的融合估计值. 实例分析表明:基于融合后的识别结果较 各传感器单独决策的结果性能优化,具有较强的容错性和有效性,并且不确定性值较融合前平均下 降了74 % ,从而可精确快速地控制工业生产.

Offered by matrix analysis of uncertainty reasoning method , a mathematical model of multisensor data fusion is estabilshed to resolve object recognition problem by analyzing comprehensively several ifferent sensor acquisitions2based normal distribution , so that basic probability assignment is easily attained according to the correlative coefficient matrix. Following this a novel data fusion algorithm is proposed. Of this approach
by the multiple levels accumulation of assignments recursively the fusion estimates based on global information are obtained.Application example shows the recognition result is optimal when compared with a single sensor ,in improving the effectiveness and correctness with an average decline 74 percent of uncertainly value . There by the imprecision and quickness of industrial production will be improved.