基于距离像时频非负稀疏编码的SAR目标识别  

SAR ATR based on HRRP time-frequency non-negative sparse coding

作　　者：张新征[1] 刘书君[1] 秦建红[1] 黄培康[2] 

机构地区：[1]重庆大学通信工程学院,重庆400044 [2]中国航天科工集团科技委,北京100854

出　　处：《系统工程与电子技术》2014年第10期1934-1941,共8页Systems Engineering and Electronics

基　　金：国家自然科学基金(61301224);中央高校基本科研业务费专项资金(CDJZR12160014;CDJRC11160003);重庆市自然科学基金(cstcjjA40018;cstc2012jjA40001)资助课题

摘　　要：提出了一种基于目标高分辨距离像时频域非负稀疏编码的合成孔径雷达(synthetic aperture radar,SAR)目标识别方法。首先,将目标的SAR复图像转换为高分辨距离像。然后,采用自适应高斯基表示方法计算每个距离像的非负时频矩阵。其次,对训练目标所有距离像的时频矩阵采用非负稀疏编码方法学习时频字典。在目标识别中,通过将每个距离像的时频矩阵投影到低维的时频字典上来提取特征矢量。最后,在提取特征矢量的基础上,通过支撑向量机目标识别决策实现目标识别。采用美国"运动和静止目标获取与识别计划"公开发布的SAR图像数据库进行算法验证实验。实验结果说明了提出方法的有效性。A new approach to classify synthetic aperture radar （SAR） targets is presented based on high range resolution profile （HRRP） time-frequency non-negative sparse coding （NNSC）. Firstly, complex SAR target images are converted into HRRPs. And the non-negative time frequency matrix for each profile is oh tained by using adaptive Gaussian representation （AGR）. Secondly, NNSC is applied to learn target time-fre- quency dictionary. Feature vectors are constructed by projecting each HRR profile time frequency matrix to the time-frequency dictionary. Finally, the target classification decision is found with the support vector machine. To demonstrate the performance of the proposed approach, experiments are performed with SAR database re leased publicly by moving and stationary target acquisition and recognition （MSTAR）. The experiment results support the effectiveness of the proposed technique for SAR target classification.

关 键 词：合成孔径雷达 自动目标识别 距离像 稀疏编码 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]