基于卷积神经网络的末敏弹复合探测信号识别方法

Recognition method of compound detection signal of terminal sensitive sub-ammunition based on convolutional neural network

为了进一步提升末敏弹的目标识别性能,提出一种基于卷积神经网络(Convolutional Neural Network,CNN)的复合探测信号识别方法。首先针对毫米波辐射计、激光测距雷达和红外敏感器的复合探测信号特点,提出了3种构造输入样本的方法;然后根据不同的信息融合方式,提出了3种基本网络架构,分别构建了单通道、多通道CNN模型对输入信号进行特征提取和分类;最后通过高塔试验数据对模型进行训练和评估。测试结果表明,基于样本构造方案2和网络结构3的识别方法表现最佳,测试准确率达到了97.26%,所提样本构造方法和识别方法能够有效提取复合探测信号的特征,具有较高的识别精度。

In order to improve the target recognition performance of terminal sensitive sub-ammunition,a compound detection signal recognition method based on convolutional neural network(CNN)is proposed.First,according to the characteristics of the composite detection signal of millimeter wave radiometer,laser ranging radar and infrared sensor,three methods of constructing input samples are proposed;then,three basic network architectures are proposed according to different information fusion methods,and single-channel and multi-channel CNN models are constructed to extract and classify the input signals;finally,these models are trained and evaluated through the high tower experiment data.The experiment results show that the recognition method based on sample construction scheme 2 and network structure 3 performs best,and the test accuracy rate reaches 97.26%.The proposed sample construction method and recognition method can effectively extract the characteristics of the composite detection signal,and have high recognition accuracy.