基于MIMO-FNN模型的弹道导弹目标时空序贯融合识别方法  

Temporal-Spatial Sequential Fusion Recognition Method of Ballistic Missile Target Based on MIMO-FNN Model

作　　者：李昌玺[1,2] 周焰 林菡[3] 李灵芝 郭戈[1] LI Changxi ZHOU Yah LIN Han LI Lingzhi GUO Ge(Air Force Early Warning Academy, Wuhan 430019, China Unit No. 66132 of PLA, Beijing 100043, China Department of Computer Science, Dongfang College, Fujian Agriculture and Forestry University, Fuzhou 350017, China)

机构地区：[1]空军预警学院 [2]中国人民解放军第66132部队 [3]福建农林大学东方学院计算机系

出　　处：《上海交通大学学报》2017年第9期1138-1144,共7页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金青年科学基金(61401503)资助

摘　　要：针对传统弹道导弹(BM)目标时空序贯融合识别算法识别效率低、抗噪性能差的缺点,提出了一种基于多输入多输出模糊神经网络(MIMO-FNN)模型的BM目标时空序贯融合识别方法.该方法首先根据多层融合的思想,结合神经网络和模糊理论,提出了多传感器多特征MIMO-FNN模型;其次,在此基础上,将当前时刻的融合结果与下一时刻的融合结果再融合,得到此时刻时空序贯融合识别结果,并将其与识别门限比较,直到满足识别门限要求,时空序贯融合识别结束,并做出决策;最后通过实验验证了所提模型的有效性和良好的抗噪性.In traditional temporal-spatial sequential fusion recognition method of ballistic missile（BM）target,there always exits the problem that the efficiency is low and the anti-noise performance is bad.In order to solve these difficulties,this paper proposed a fusion recognition method which is called temporalspatial sequential fusion recognition method of BM target based on multiple input multiple output and fuzzy neural network（MIMO-FNN）model.In this model,firstly,we use the idea of multi layer fusion,combine with neural network and fuzzy theory,and put forward the MIMO-FNN model with multi sensors of multi features.And then,we reintegrate the results of present moment and next moment to get the fusion results.Meanwhile,we compare with recognition threshold,until the fusion results match the recognition threshold,and the process of fusion ends.Finally,the experiment validates the effectiveness and antinoise performance of this model.

关 键 词：多输入多输出 模糊神经网络 弹道导弹 时空序贯融合 目标识别 

分 类 号：TN95[电子电信—信号与信息处理]