基于插值HRRP和SVM的雷达目标识别方法  

Radar Target Identification Method Based on the Interpolated HRRP and SVM

作　　者：梁海涛[1] 童创明[1] 王晓丹[1] 郑颖[2] 

机构地区：[1]空军工程大学导弹学院,陕西三原713800 [2]西北工业大学软件与微电子学院,西安710053

出　　处：《弹箭与制导学报》2008年第3期210-212,226,共4页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：陕西省自然科学研究计划项目(2004F36,2005F23)资助

摘　　要：为了在高分辨一维距离像(HRRP)数据有限的情况下提高支持向量机(SVM)的识别性能,采用插值方法对原数据按方位角进行插值处理。给出了应用于距离像数据的线性插值和三次样条插值方法的推导和计算过程以及处理结果,与未插值数据的图像吻合较好,证明了插值方法的正确性。研究采用SVM方法对插值数据及未插值数据进行识别,结果表明采用插值处理可明显提高SVM的正确识别率,采用三次样条插值比线性插值提高识别率的效果更明显,且三次样条插值在两点间插入5个点后对识别率的提高可趋于稳定,从而证明该方法的正确性和有效性。In order to improve the identification performance of support vector machine（SVM）under the condition of limited high resolution range profile（HRRP）, the interpolating method was studied and adopted to interpolate the HRRP along azimuth. The reasoning, calculating process and interpolated results of linear and thrice spline interpolating methods were presented, and the results matched original map well and the correctness of interpolating method was proved. The SVM method was studied and adopted for identifying the original HRRP and interpolated HRRP. The identifying result indicates the interpolating methods can greatly improve the correct identification rate of SVM , and the thrice spline method can better improve the correct identification rate than the linear, and the improved correct identification rate can be stable when the thrice spline interpolating five or more data-node between two original data. So, the validity and correctness of this method were proved.

关 键 词：雷达目标识别 支持向量机 插值 高分辨一维距离像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN95[自动化与计算机技术—计算机科学与技术]