传感器可靠性相异的信任函数理论融合识别算法研究

Research on fusion recognition algorithm for different reliable sensors Based on the belief function theory

本文在信任函数理论框架下,针对决策层融合目标识别问题,对可靠性相异的多传感器融合目标识别提出了一种新算法。该算法结合各传感器混淆矩阵的先验静态信息以及其当前输出判决的动态信息,获得各传感器当前输出识别证据的可靠性因子,并将其用于相应信任函数的折扣,最后用Dempster组合规则对修改后的识别证据进行融合。仿真表明和现有一些考虑可靠性问题的融合算法相比,本文算法能够得到更好的融合效果。

Under the framework of belief function theory, a new fusion recognition algorithm for different reliable sensors was proposed at decision level. Combining the prior static information of each sensor＇ s confusion matrix with the dynamic information of its current decision, the algorithm acquires the reliability factor of each sensor＇ s current output recognition evidence. These reliability factors are applied to the discount of corresponding evidence and then these modified evidences are combined by Dempster＇ s combination rule. The simulation experiment indicates the algorithm outperforms these exiting fusion methods which takes the reliability into account.