基于1D-CNN 的弹链运动加速度分类与识别

针对人工判读研究弹链运动规律时存在过程复杂、效果不佳的问题,结合弹链运动加速度的1维特性,提出一种基于1D-CNN的弹链运动加速度分类与识别方法。基于Keras深度学习框架搭建1维卷积神经网络模型(1D convolutional neural network,1D-CNN),对小口径自动炮射击试验中获取的弹链运动加速度信号进行数据预处理并制作训练集和测试集,利用训练集和测试集对1D-CNN模型进行训练和测试。结果表明:利用1D-CNN模型可实现弹链运动加速度信号的分类和识别,准确率在84%左右,达到了预期效果。

Classification and Recognition of Ammunition Link Motion Acceleration Based on 1D-CNN

Aiming at the problems of complex process and poor effect in manual interpretation of ammunition chain motion law, a classification and recognition method of ammunition chain motion acceleration based on 1D convolutional neural network (1D-CNN) model is proposed by combining the one-dimensional characteristics of ammunition chain motion acceleration. The 1D-CNN model was built based on Keras deep learning framework. The data of ammunition chain motion acceleration signal obtained from small bore automatic gun firing test was preprocessed, and the training set and test set were made to train and test the 1D convolutional neural network model. The results show that the 1D-CNN model can realize the classification and recognition of ammunition chain motion acceleration signal, and the accuracy rate is about 84%, which achieves the expected effect.