基于Brushlet和自适应脊波网络的SAR图像分类  

SAR Image Classification Based on Brushlet and Self Adaptive Ridge-Wavelet Neural Network

作　　者：童涛[1] 杨桄[1] 李昕[2] 叶怡[1] 王寿彪[1] 

机构地区：[1]空军航空大学特种专业系,长春130022 [2]空军航空大学训练部,长春130022

出　　处：《兵工自动化》2012年第9期67-70,共4页Ordnance Industry Automation

基　　金：国家自然科学基金项目(40901096)

摘　　要：针对BP神经网络分类模型网络训练时间长的缺陷,结合小波网络模型缺少对方向信息描述的情况,提出一种基于Brushlet和自适应脊波网络的SAR图像分类方法。该方法利用Brushlet变换提取SAR图像纹理的能量和相位特征,并将描述能量和相位的特征向量输入到自适应脊波网络中进行训练和分类,最后通过实验对比分析各分类方法的分类性能。实验结果表明,该方法快速、准确,其性能优于传统方法。Aiming at the defect that BP neural network classification model taking a long time for network training, combine with the condition that wavelet network model＇s shortness of direction information depicting, put forward a method for SAR image classification, which uses the energy and phase feature of Brushlet as texture feature of SAR image, in addition, inputs the feature vector which describes energy and phase information to adaptive ridge-wavelet neural network for training and classifying. At last, compare and analyze the classification features through test the experiment results on SAR image show that this method is rapid and accurate and outperforms the traditional methods.

关 键 词：SAR图像分类 Brushlet复特征 自适应脊波网络 纹理图像 

分 类 号：TJ03[兵器科学与技术—兵器发射理论与技术]