合成孔径雷达图像的分类方法研究仿真  

Simulation Study on Method of Synthetic Aperture Radar Image Classification

作　　者：冯东栋[1] 张钊 

机构地区：[1]河南大学软件学院,河南开封475004 [2]河南省医药学校,河南开封475000

出　　处：《计算机仿真》2013年第9期382-385,共4页Computer Simulation

摘　　要：合成孔径雷达有多个通道,采集得到的远程雷达图象特征较多,图像合成后由于特征的不一致,造成图像维度很高。传统的图像分类方法在设置训练器时,一旦雷达图像特征维数很高,会使得传统的分类器难以训练至给定要求,模型的分类中心发生偏移,造成图像分类不准。提出了一种遗传因子优化的超高维度合成孔径雷达图像分类算法。利用遗传因子优化方法,对采集的雷达图像分类参数进行优化训练,从而获取最优参数。利用搜索空间划分方法,对采集的雷达图像进行分类处理。实验结果表明,改进的算法进行高精度合成孔径雷达图像分类,能够有效提高分类的准确性,从而为航天技术的发展提供依据。Synthetic aperture radar itself has multiple channels,more characteristics of long range radar image are acquired after image synthesis,which causes high image dimension due to the inconsistent characteristics.A synthetic aperture radar image classification algorithm in high dimension based on genetic factor optimization was proposed.The optimization method of genetic factors was utilized to optimize the acquired parameters of radar image classification training,so as to obtain the optimal parameters.A search space partition method was employed to classify the acquired radar image.The experimental results show that the optimized algorithm used in high precision synthetic aperture radar image classification,can effectively improve the accuracy of classification,so as to provide the basis for the development of space technology.

关 键 词：合成孔径 雷达图像 图像分类 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]