基于超分辨ISAR成像的飞机目标SVM分类算法

The SVM Classification Method for Airplane Target based on Super-resolution ISAR Imaging

利用最大熵谱估计方法对4种飞机目标数据进行外推处理,并在此基础上进行拟合成孔径(ISAR)成像。采用了ISAR图像的几何矩、基于几何矩的不变量、形状和量化能量带4个特征,研究了支持向量机的线性和非线性算法原理,提出了基于SVM的飞机目标识别和分类算法,采用了针对多目标分类的M-ary法对飞机进行分类,选取了每个目标的40个不同数据段进行成像,通过与几种常见的BP神经网络算法和RBF神经网络算法比较分析,验证结果表明此方法达到了较好的识别效果,识别率达到97%。

The method of Maximum Entropy Spectrum Estimation is used to extrapolate four airplane data, and ISAR imaging is done on the basis of it. The four characters of ISAR images, i.e. geometric moment, invariants based on geometric moment, shape and quantized energy belt are adopted in the study of the arithmetic theory of linearity and non - linearity about SVM, and a method of target recognition and classification based on SVM is pro- posed. The M - ary method for multi - target classification is used to classify the airplanes, and 40 different data fields of each target are chosen for imaging. Compared with the usual BP and RBF neural network algorithms, this method is good in recognition, and its discrimination can reach 97%.