基于稳健深层网络的雷达高分辨距离像目标特征提取算法  

Feature Extraction Method for Radar High Resolution Range Profile Targets Based on Robust Deep Networks

作　　者：冯博[1] 陈渤[1] 王鹏辉[1] 刘宏伟[1] 

机构地区：[1]西安电子科技大学雷达信号处理国家重点实验室,西安710071

出　　处：《电子与信息学报》2014年第12期2949-2955,共7页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61372132;61201292);新世纪优秀人才支持计划(NCET-13-0945);中央高校基本科研业务费专项资金资助课题

摘　　要：特征提取是雷达高分辨距离像(HRRP)目标识别的核心技术。传统的特征提取算法多采用浅层的模型结构,容易忽视样本的内在结构,不利于学习有效的分类特征。针对这一问题,该文利用多层非线性网络实现特征学习,构建了基于深层网络的雷达HRRP目标识别框架。利用平均像在散射点不发生越距离单元走动的方位帧内具有稳健物理特性的性质,提出了堆栈联合稳健自编码器。该网络由一系列联合稳健自编码器堆栈化实现,在匹配原始HRRP样本的同时,约束同帧样本趋近于平均像,并将网络的最终输出作为分类器的特征输入。基于实测HRRP数据的实验结果验证了所提算法的有效性。Feature extraction is the key technique for Radar Automatic Target Recognition （RATR） based on High Resolution Range Profile （HRRP）. Traditional feature extraction algorithms usually use shallow models. When applying such models, the inherent structure of the target is always ignored, which is disadvantageous for learning effective features. To address this issue, a deep framework for radar HRRP target recognition is proposed, which adopts multi-layered nonlinear networks for feature learning. Ground on the stable physical properties of the average profile in each HRRP frame without migration through resolution cell, Stacked Robust Auto-Encoders （SRAEs） are further developed, which are stacked by a series of RAEs. SRAEs can not only reconstruct the original HRRP samples, but also constrain the HRRPs in one frame close to the average profile. Then the top-level output of the networks is used as the input to the classifier. Experimental results on measured radar HRRP dataset validate the effectiveness of the proposed method.

关 键 词：雷达自动目标识别 高分辨距离像 深层网络 堆栈联合稳健自编码器 

分 类 号：TN957.51[电子电信—信号与信息处理]