无监督神经网络的潜艇对空战术意图识别

摘要

The traditional way to obtain the air target information through the remote sensing system so as to resolve the target tactical intent requires a large number of experts to evaluate the network nodes and weights with a slow speed,high cost and other shortcomings.To reduce the recognition time of air combat,the computing ability of unsupervised learning neural network is brought into full play.The air target attribute and target tactical intent acquired from remote sensing are used to form the training samples to train the neural network,and thus the input threshold target attribute and relationship between neurons in competitive layer are acquired.The output function is established,and the air target tactical intention is identified.The simulation results show that the output value of the test sample trained by the competitive neural network and self-organizing feature maping (SOFM) neural network corresponds to the real value,and the accuracy is higher.%传统的通过遥感系统获取空中目标信息,解析出目标战术意图的方式,需要大量的专家评估计算网络节点及权重,具有速度慢,耗费大等缺点.为了降低潜艇对空战术意图识别时间,发挥无监督学习神经网络的计算能力.利用遥感获取的空中目标属性与目标战术意图形成训练样本训练神经网络,获得输入目标属性的阈值及竞争层神经元间的关系,建立输出函数,识别空中目标的战术意图.仿真结果表明,竞争神经网络与自组织特征映射(SOFM)神经网络训练的测试样本的输出值与真实值相对应,准确度较高.