摘要
本文在分析传统二维鉴别分析方法局限性的基础上,提出了一种基于二维子分类鉴别分析的合成孔径雷达图像识别方法.该方法首先对SAR图像进行图像预处理,然后利用图像欧氏距离对每类目标进行子类划分,并由图像的行信息和列信息提出了两种二维子分类鉴别分析方法,最后利用最近邻分类器对提取的特征投影矩阵进行分类识别.本文利用美国实测的MSTAR数据对算法进行了仿真验证,实验结果表明了本文方法的正确性和有效性.
By analyzing the limitation of the traditional two-dimensional LDA,a new SAR image recognition algorithm based on two-dimensional subclass discriminant analysis is proposed in this paper.First,image preprocessing of SAR image is performed,which includes image aligning and power transformation.Then image Euclidean distance NN-Clustering is proposed to divide the datasets per class into multi-subclasses,and feature matrix is extracted by two new methods,i.e.2DSDA and Alternative-2DSDA.Finally,a nearest neighbor classifier is employed to classify the extracted features.Experimental results with MSTAR dataset verify the correctness and effectiveness of the proposed method.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2010年第4期798-803,797,共7页
Acta Electronica Sinica
基金
全国优秀博士学位论文作者专项基金(No.20043)
"泰山学者"建设工程专项项经费
关键词
SAR
自动目标识别
二维子分类鉴别分析
图像欧氏距离
特征提取
synthetic aperture radar（SAR）
automatic target recognition
two-dimensional subclass discriminant analysis（2DSDA）
image Euclidean distance（IED）
feature extraction
