多分辨率特征在UWB SAR目标检测中的应用  

Application of multi-resolution feature in UWB SAR target detection

作　　者：杨志国[1] 黄晓涛[1] 周智敏[1] 

机构地区：[1]国防科学技术大学电子科学与工程学院超宽带室,湖南长沙410073

出　　处：《信号处理》2008年第3期386-389,共4页Journal of Signal Processing

基　　金：国家自然科学基金(60402034)

摘　　要：本文借鉴高频合成孔径雷达(SAR)目标检测中多分辨率特征分析方法,结合超宽带合成孔径雷达(UWB SAR)工作体制和树干杂波的特殊性,详细分析了UWB SAR目标检测中树干杂波和目标多分辨率特性的区别;构造了多分辨率差分序列,并采用一阶AR模型实现了对多分辨率差分序列的建模;给出了基于多分辨率的广义似然比(GLR)定义,并提出了利用感兴趣区域(ROI)内GLR之和来确定ROI真假属性的判别方法;针对树干杂波和目标的模型训练分别提出了预先训练和现场训练的方法,大大提高了多分辨率特征提取的稳健性;基于实际UWB SAR图像数据的试验结果证明了利用多分辨率特征降低ROI虚警率的有效性。The paper detailedly analyzes the difference of multi-resolution properties between trunk clutter and targets in ultra- wide band （UWB） synthetic aperture radar （SAR） target detection based on the particularities of operation system of UWB SAR and trunk clutter consulting the analysis of multi-resolution feature in the target detection of high-frequency SAR. We construct the difference sequences and use a first-order autoregression （AR） to model the difference sequences. The paper gives the definition of generalized likelihood ratio （GLR） based on multi-resolution and presents a judgment approach of region of interest （ROI） property using the summation of GLR in a ROI. We present a method of advance training and a method of local training respectively for the model training of trunk clutter and target that improves the steadiness of multi-resolution feature extraction. The experimental results based on the actual image data of UWB SAR testify that the multi-resolution feature can be used to reduce the false alarm probability of ROI effectively.

关 键 词：UWB SAR 检测 ROI 多分辨率 差分序列 一阶AR模型 GLR 

分 类 号：TN957.51[电子电信—信号与信息处理] TP391.41[电子电信—信息与通信工程]