论文部分内容阅读
本文讨论了基于径向基函数网络(RBFN)的雷达目标识別问题。在分析了一维距离象特点的基础上,提出了采用非相关幅度平均一维距离象以获取稳定模式这一有效方法。在指出传统经验公式局限性后,给出了一种基于训练样本空间分布来估计高斯核函数形状参数的方法。用微波暗室试验数据进行转台成象并对一维距离象三种模式进行识别分类的结果表明,本文所提出的方法用于研究雷达目标识别是有效的。
This paper discusses the radar target recognition problem based on radial basis function network (RBFN). Based on the analysis of the characteristics of one-dimensional distance images, an effective method of using non-correlated amplitude-averaged one-dimensional distance images to obtain a stable mode is proposed. After pointing out the limitation of the traditional empirical formula, a method of estimating the shape parameters of Gaussian kernel function based on the spatial distribution of training samples is given. The results of the turntable imaging using microwave anechoic chamber test data and the recognition of one-dimensional distance from the three modes show that the proposed method is effective for studying radar target identification.