文献类型：期刊文献

中文题名：跨卷积网络特征融合的SAR图像目标识别

英文题名：SAR Image Target Recognition Based on Across Convolution Network Feature Fusion

作者：冯新扬[1];邵超[1]

第一作者：冯新扬

机构：[1]河南财经政法大学计算机与信息工程学院,河南郑州450046

第一机构：河南财经政法大学计算机与信息工程学院

年份：2021

卷号：33

期号：3

起止页码：554-561

中文期刊名：系统仿真学报

外文期刊名：Journal of System Simulation

收录：CSTPCD;;Scopus;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家自然科学基金(61202285)。

语种：中文

中文关键词：合成孔径雷达;Le Net-5神经网络;协作表示分类;深层特征

外文关键词：synthetic aperture radar;LeNet-5 neural network;collaborative representation classification;deep feature;

摘要：卷积神经网络(Convolutional Neural Network,CNN)在合成孔径雷达(Synthetic Aperture Radar,SAR)图像目标识别领域得到广泛应用。在Le Net-5神经网络模型的基础上,提出了跨卷积网络特征融合的SAR图像识别方法。利用MNIST手写数据对LeNet-5网络参数进行初始化,提取SAR图像的深层特征和浅层特征,对浅层特征进行主成分分析以得到关键类别信息,将深层特征和浅层特征进行融合,使用协作表示分类(Collaborative Representation Classification, CRC)将融合的两部分进行识别。通过公开数据集的实验验证表明,在不扩充训练样本条件下,该方法可达到98%的平均识别率。
Convolutional neural networks have been widely used in the field of synthetic aperture radar image target recognition.Based on the LeNet-5 neural network model,a SAR image target recognition method are initialized across convolution network feature fusion is proposed.The LeNet-5 network parameters on the basis of MNIST handwritten data.The deep and shallow features of the SAR image are extracted,and the principal component analysis on the shallow features is performed to obtain key category information.Deep features and shallow features are fused and are classified and recognised by sent to collaborative representation.Experimental results show that the method can achieve 98%average recognition rate without expanding the training samples.