小样本下基于CNN-GRU网络的弹丸落点预测
Projectile Impact Point Prediction Based on CNN-GRU Network under Small Samples
王现磊 1陈铎 1薛景元 1王义江1
作者信息
- 1. 解放军63861部队,吉林 白城 137001
- 折叠
摘要
为充分挖掘弹丸径向速度在时空上的规律,提高弹丸落点预测精度,提出了一种基于卷积神经网络结合门控循环单元的弹丸落点预测方法.分别利用卷积神经网络(CNN)和门控循环单元(GRU)网络对弹丸的径向速度在时间和空间上的强相关性特征进行提取,学习弹丸高度复杂的非线性飞行轨迹,构建弹丸落点预测模型.通过某型炮弹径向速度数据作为训练集和测试集进行弹丸落点预测,并与MLP、LSTM和CNN-LSTM时序预测方法进行比较.实验结果表明,CNN-GRU预测模型能够有效提取径向速度序列中的时空间信息,学习出弹丸相对雷达的位置,相比其他模型具有预测精度高、收敛速度快且稳定性好的优势.
Abstract
In order to fully explore the law of projectile radial velocity in time and space,and im-prove the accuracy of projectile impact-point prediction,a method of projectile impact point prediction based on CNN-GRU is proposed.CNN and GRU networks are used respectively to extract the strong correlation characteristics of the projectile radial velocity in time and space,to learn the highly complex nonlinear flight trajectory,and to build the prediction model of projectile impact points.The radial ve-locity data of a certain type of projectile is used as the training set and test set to predict the impact points,and compared with the time series prediction methods of MLP,LSTM and CNN-LSTM.The ex-perimental results show that the CNN-GRU prediction model can effectively extract the spatiotemporal information in the projectile radial velocity sequence,and learn the position of the projectile relative to the radar.The comparisons with other models show that the predication model has higher prediction ac-curacy,faster convergence speed and better stability.
关键词
径向速度/落点预测/卷积神经网络/门控循环单元Key words
radial velocity/impact-point prediction/convolutional neural network/gated recurrent unit引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据