队形识别智能化方法定量分析  

Quantitatively Analysis on Formation Identification Intelligent Method

作　　者：田立业[1] 刘剑[1] 鲍凯[1] TIAN Li-ye;LIU Jian;BAO Kai(Navy Submarine Academy,Qingdao 266100,China)

机构地区：[1]海军潜艇学院

出　　处：《指挥控制与仿真》2019年第6期18-23,共6页Command Control & Simulation

摘　　要：潜艇战术软件对于队形识别功能有一定需求,队形识别本质上属于模式识别问题,用智能化方法实现队形识别已经得到了定性论证。对四种人工神经网络方法进行了定量试验,为了精确测试试验基于小样本进行,样本通过0和1标记的方式获取,试验了横队、纵队、人字队等多个不同测试集,增加反人字队形后的情况也进行了对比。GRNN网络在四次试验中表现出良好的泛化能力和队形识别能力,可用于小样本情况下的队形识别。Submarine tactical software needs formation identification ability.It essentially belongs to pattern recognition issue.Artificial intelligence methods application has already get qualitative demonstration.This paper quantitatively tests four artificial neural networks.In order to carry out precise tests,the test is setup based on small samples.Those samples are collected by marking 0 and 1.Various samples are tested.Three formations and four formations cases are both tested and compared.In these 4 tests,GRNN network shows good generalization ability and formation identification ability.It works well in small sample formation identification.

关 键 词：潜艇战术软件 队形识别 神经网络 广义回归网络 模式识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] E917[自动化与计算机技术—计算机科学与技术]