雷达目标识别中的维数压缩

在雷达目标识别中,维数压缩(降维)的主要目的不仅是要降低特征量的维数、减少计算量,同时还要尽可能地保留分类鉴别信息,增加不同类别数据之间的可分性,提高目标的识别概率.K-L变换(KLT)是一种常用的数据压缩方法,但它只是按照均方误差(MSE)最小的原则来保留信息,本文在相对KL变换(RKLT)和多空间KL变换(MKLT)的基础上给出了一种新的压缩变换矩阵形式,它不仅能够使均方误差达到最小,而且可以提高识别概率.

Dimensionality Reduction in Radar Target Identification

In radar target identification, dimensionality reduction is not only used for reducing the dimension of the feature vector and the burden of the computation, but also remains the classification information and increases the classification ability of various classes to improve the identification probability of the target. Karhunen-Loeve transform (KLT) is a well known method for dimensionality reduction, but it only preserves information about the inputs in the sense of minimizing the mean square error (MSE). A new compression matrix transform method is presented based on RKLT and MKLT. When it is used for high resolution radar range profile, it not only satisfies the minimization of MSE, but also improves the identification probability.