基于神经网络的目标分群及目标运动状态预测  

Target Grouping and Target Motion State Prediction Based on Neural Network

作　　者：张永利 刘楠楠 王兆伟 ZHANG Yong-li;LIU Nan-nan;WANG Zhao-wei(China Academy of Electronics and Information Technology,CETC,Beijing 100041,China)

机构地区：[1]中国电子科技集团公司电子科学研究院,北京100041

出　　处：《舰船电子对抗》2020年第3期7-12,共6页Shipboard Electronic Countermeasure

摘　　要：态势估计是对动态变化的对象感知并对提取的态势要素进行觉察、理解和预测的处理过程。以目标运动状态特征为依据,基于自组织竞争神经网络,对传感器量测数据进行无监督的自组织聚类,将各个量测数据准确划分到各个类别当中,解决目标分群的问题。对于具有明显的时变性和不确定性的空中目标的运动特征,Elman神经网络可反映系统随时间变化的动态特性及存储信息的能力,预测任意逼近非线性函数,通过学习历史数据建模,对目标的运动特征进行网络训练,预测目标运动状态,为实现多目标环境下的航迹接续提供方法借鉴。Situation assessment is the process to sense the dynamic changing objects,and makes the perceivation,apprehension and prediction to the extracted situation factors.According to the features of object motion state,unsupervised self-organizing clustering is carried on to sensors’measurement data based on self-organizing competitive neural network.Measurement data can be divided into various categories to solve the problem of target grouping.To the motion features of air targets with obvious time-varying and uncertainty,the Elman neural network can reflect the dynamic property that the system changes with time and the information storage capability,predict random approaching non-linear functions,model by learning history data,perform the network training to object motion features,predict object motion states,which can provide the reference for the track continuation under the multiple targets environment.

关 键 词：自组织竞争神经网络 目标分群 ELMAN神经网络 目标运动状态预测 态势估计 

分 类 号：TN971.1[电子电信—信号与信息处理]