决策层信息融合的神经网络模型与算法研究

本文对信息融合问题中决策层融合方法进行了分析与比较,提出了一种新的决策层信息融合算法,即改进型ＡＲＴ２神经网络融合算法,该融合算法在综合大脑对多源信息融合的特点和优势基础上,提出了将信息进行匹配和调和相融合的处理方式。对实际的决策层信息融合目标识别问题,该算法具有弹性去除信息间相关性以及合理处理矛盾信息的能力。同时,ＭＡＲＴ神经网络模型通过自适应地调整网络参数,对信度的增长有较好的控制能力。

Decision Fusion for Target Recognition Based on Neural Network Model

Compared with normal decision fusion methods,a novel decision fusion algorithmbased on modified ART2(MART)Peural network model is presented in this paper. It has advantages of human brain,and the algorithm uses "matching function"and "compromising function" toremove the correlation of information and to obtain the ability of processing contradictory infromation. Meantime,the MART model can control the output of belief-level by adaptively adjusting theparameters of network.