摘要
海上舰艇防空反导作战基于目标攻击意图识别是现代舰艇防空技术的研究热点;来袭目标攻击意图识别是战场态势分析的一个重要部分,以往是通过先验知识和先验概率进行量化分析与明确攻击意图识别特征值的影响权重;深度神经网络可通过自适应学习目标攻击意图的特征值,可以在缺乏先验知识的条件下,通过小样本集的目标攻击意图的特征值训练,学习特征数据和攻击意图识别之间的对应关系与映射;通过引入GeLUs激活函数和AdaMod优化算法加快模型收敛,并解决了Adam模型可能无法收敛到最优解的问题;实验结果显示文中提出的模型可以在先验知识不足及训练数据规模小的情况下,有效识别来袭目标攻击意图,同时保证更高的准确率。
Naval ship air defense and antimissile operations based on target attack intention recognition are research hottopics in naval ship air defense technology.The attack intention identification of attacking targets is an important part of battlefield situation analysis.In the past,quantitative analysis is conducted through prior knowledge and probability,and the impact weights of attack intention identification feature values are determined.Deep neural networks can adaptively learn the feature value of target attacking intentions,and learn the correspondence and mapping between feature data and attacking intention recognition through the feature value training of target attacking intentions in small sample sets without prior knowledge.The Gaussian error linear units(GeLUs)activation function and adaptive and momental bound(AdaMod)optimization algorithm are introduced to accelerate the convergence of the model,and solve the problem that the Adam model may not converge to the optimal solution.Experimental results show that the proposed model can effectively identify the attack intent of attacking targets with insufficient prior knowledge and small training data,while ensuring higher accuracy.
作者
王家鑫
王瑞琪
孟海波
蔺红明
陈天群
WANG Jiaxin;WANG Ruiqi;MENG Haibo;LIN Hongming;CHEN Tianqun(Haizhuang is Stationed in a Department in Shanghai,Shanghai 201109,China;Shanghai Institute of Mechanical and Electrical Engineering,Shanghai 201109,China)
出处
《计算机测量与控制》
2023年第6期274-279,共6页
Computer Measurement &Control
