针对合成孔径雷达图像数据源的分类优化方法,提出了基于多层卷积神经网络的 SAR 图像分类方法.该方法采用多层卷积运算和下采样技术以及神经元的非线性功能,利用局部响应归一化进行特征的降维,以 softmax作为分类器.用 MSTAR数据库的五类目标数据进行仿真实验,仿真实验结果表明该方法是有效的,统计平均识别率达到了 97.77%.%In order to improve the classification and optimization of synthetic aperture radar (SAR)image data source,a SAR image classification method based on multilayer convolution neural network was proposed.The method used multi-layer convolutional operation and downsampling techniques as well as the neuron's nonlinear function.The local response was normalized to reduce the dimension of features,and softmax was used as the classifier.Experimental results showed that the proposed method was effective and the average statistical recog-nition rate reached 97.77% with five kinds of target data of MSTAR database.
展开▼