基于神经网络和DS证据理论的多传感器数据融合目标识别方法

A multi-source data fusion target recognition method based on neural network and DS evidence theory

针对战场态势感知多传感器数据的目标识别问题,提出基于神经网络和DS证据理论相结合的数据融合目标识别方法。首先,充分发挥神经网络强非线性映射、自学习和自适应的能力,获取单个传感器中数据之间的关联关系,完成特征级融合的分类输入/输出;然后,根据混淆矩阵获取各传感器的全局可信度和局部可信度,实现基本概率分配函数的赋值;最后,利用DS证据理论对不同数据源所提供的局部识别能力加以整合,降低不确定性,完成多传感器数据融合的决策输入/输出。实验结果表明:提出的方法可有效融合不同传感器信息,最大限度地提高目标识别判断的置信度。

Aiming at the target identification problem of multi-sensor data in battlefield situational awareness,a multi-source data fusion method based on the combination of neural network and Demister-Shafer(DS)evidence theory is proposed in this paper.Firstly,the strong non-linear mapping,self-learning and self-adaptive capabilities of neural network are fully utilized to obtain the correlation between the data in each sensor,and complete the classification input/output of feature-level fusion.Then,the global credibility and local credibility of each data sensor are obtained according to the confusion matrix,and the assignment of the basic probability distribution function is interned.Finally,the DS evidence theory is used to combine the local incomplete observations provided by different data sources to reduce uncertainty and complete the decision input/output of multi-sensor data fusion.The experimental results show that the proposed method can effectively fuse sensor information and maximize the confidence level of target identification judgement.