基于目标区域匹配的SAR目标识别方法  

SAR Target Recognition Based on Target Region Matching

作　　者：付凡成[1] FU Fan-cheng(School of Computer Information Engineering, Nanchang Institute of Technology, Nanchang 330044, Chin)

机构地区：[1]南昌理工学院计算机信息工程学院,南昌330044

出　　处：《电光与控制》2018年第4期37-40,共4页Electronics Optics & Control

基　　金：江西省级教改课题(JXJG-2011-687)

摘　　要：特征提取和特征匹配是合成孔径雷达(SAR)目标识别中的两个关键步骤。提出了一种基于SAR目标区域匹配的目标识别方法。首先提取SAR图像中二值化目标区域;然后将其与模板库中对应的目标区域作差得到残差图像,采用欧氏距离变换对残差图像进行处理;最后利用距离变换后的残差图像构建相似度度量标准,计算当前待识别图像与各类目标的匹配度并根据最大匹配度原则判定目标类型。目标区域残差可以体现待识别目标与其他类目标之间物理尺寸的差异,因此可以根据残差的面积大小以及形状分布计算匹配度。欧氏距离变化可以较好地体现出目标区域残差的形状分布特性。基于欧氏距离变换后的残差可以更有效地反映目标区域的匹配度。采用MSTAR数据集进行了目标识别实验,验证了方法的有效性。Feature extraction and feature matching are two key steps in SAR target recognition. A method for SAR target recognition based on the matching of target regions is proposed. First, the binary target region of SAR image is extracted, which is then compared with the corresponding target region in the template base to produce the residual image. Euclidean distance transform is used to process the residual image, and the processed image is used for constructing a similarity criterion. The matching rate of the image to be identified with each type of target is then calculated out, and the type of target is judged according to the principle of the maximum matching rate. The target region residual can reflect the physical difference of the target to be identified with the other types of targets, and thus we can calculate the matching rate by use of the area and shape distribution of the residual. The change of Euclidean distance can reflect the shape distribution of the target region residual, and the residual based on the Euclidean distance transform can better reflect the matching rate of the target region. Target recognition test was carried out by using MSTAR dataset, and the result verifies the effectiveness of the method.

关 键 词：合成孔径雷达 目标识别 目标区域 区域匹配 欧氏距离变换 特征提取 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]