摘要
提出了一种有效的高分辨率SAR目标特征提取与识别方法,根据SAR图像目标对多尺度Gabor滤波器组的不同响应,充分利用多尺度信息及尺度间的相依性提取新的多尺度特征,该特征综合考虑了SAR图像目标的宏观和微观固有的尺度特性,从而更能反映目标的本质特征;并利用Fisher核映射使得非线性变换比线性变换更能刻画实际中的复杂模式。实验给出了MSTAR数据库中三类目标特征的空间聚类及分类情况,结果表明了该多尺度特征的有效性。
An effective method for high resolution synthetic aperture radar（SAR） images feature extraction and target recognition is presented.According to the different response characters of targets to multi-scales Gabor filters,multi-scale information and the pertinence between multi-scales are adopted to extract new multi-scale features.This new feature combines target＇s macroscopical and microcosmical characteristic together,that makes it easy to obtain more essential features of SAR image.Nonlinear transform that adopts kernel Fisher is more capable of describing complex patterns than linear transformation.The experimental results show the validities of this new multi-scale feature.
出处
《武汉大学学报（信息科学版）》
EI
CSCD
北大核心
2010年第8期946-950,共5页
Geomatics and Information Science of Wuhan University
基金
国家自然科学基金资助项目(40871209)
