雷达一维距离像目标识别方法性能的研究

Performance study of target recognition methods based on the 1D HRR profile

给出了雷达距离像的数学模型,从平均距离像和距离像的协方差阵出发,用判别距离观点分析了常用的基线法、相关法、双谱法、主本征向量法和二次型法的分类性质.用判别距离准则将它们分为两类:基于平均距离像的线性分类器,基于平均距离像和协方差阵的二次型分类器.分析结果表明,二次型分类器利用了多个距离像的协方差信息,比线性分类器具有更好的分类效果.

The mathematical model of HRR profiles is first given, and then the classification performance of the existing methods such as the Baseline method, Correlation method, Bispectrum method, Principal Eigenvector method, and Quardratic Classification method is analyzed by the distance discriminant criterion, from the phase of the averaged HRR profile (AHHRP) and the covariance matrix (COVM) of the profiles. The above methods are divided into two categories: linear classifiers based on AHHRP, and quadratic classifiers based on both AHHRP and COVM. The obtained result shows that the quadratic classifier has better performance in classification than the linear one because it uses the information in COVM of profiles besides AHHRP.