基于孪生网络的机载相控阵雷达工作状态识别方法

Identification Method of Airborne Phased Array Radar Working State Based on Siamese Network

实际作战中,对敌方战机多功能相控阵雷达的工作模式状态识别结果对于飞行员的空中技战术决策至关重要。提出一种基于孪生神经网络的雷达工作状态识别方法,首先按照一定时间窗将雷达脉冲采样序列转化为二维图像,然后构造孪生网络对二维图像数据进行训练,生成在线可用的神经网络识别模型。考虑到机载相控阵雷达可能具有多种功率管理措施,以及在复杂电磁环境及干扰条件下己方接收到的雷达有效脉冲序列存在着不同程度的脉冲丢失,在训练前使用了图像数据增强方法。经仿真数据验证,所提方法对机载相控阵雷达具有较高的状态识别正确率,在脉冲丢失70%的情况下正确率仍优于90%,表明了该方法的有效性。

In actual combat,work state recognition of airborne phased array radar is crucial for pilot’s technical and tactical decision.A work state recognition method based on siamese neural network is proposed.Firstly,the pulse sequence of target radar is converted into two-dimensional image according to a fixed time window.Then,a siamese network is constructed and the network with the two-dimensional image data is trained.Considering that the target radar has a variety of power management configurations,and at the same time in a complex electromagnetic environment,the received target radar effective pulse sequences have different degrees of pulse loss,data augmentation is performed on the two-dimensional image data before training.The simulation results show that this method has a good accuracy for airborne phased array radar’s work state recognition.The accuracy is better than 90%even with 70%pulse loss,which proves the effectiveness of the proposed method.