Abstract: As for high resolution radar target recognition, the classification performance depends on feature extraction and classifier designing. To solve the problem of sensitivity characteristics of HRRP, sequential preprocessing method is taken, which enhances the signal-to-noise ratio effectively. Some features such as general central moments and distribution entropy of HRRP are extracted to form a multi-dimensional feature vector which can describe the scattering property of target better. A Back-Propagation（BP） neural network classifier is designed. A method for high resolution radar target recognition based on multi-
dimensional features and BP neural network is proposed. The measured echoes data samples in the anechoic chamber are processed by means of the BP neural network classifier to discriminate three kinds of target from each other. Experimental results demonstrate that the method can classify targets with performances of over 95% correct classification rate and less than 3% false alarm rate.

Key words: target recognition, High Resolution Range Profile（HRRP）, sequential preprocessing, Back Propagation（BP） neural network, multi-dimensional feature vector

摘要： 高分辨雷达目标的识别性能取决于目标特征的提取以及分类器的设计。为解决雷达高分辨距离像（HRRP）的方位、平移和幅度敏感性问题，采用了序贯预处理方法，有效提高了HRRP的信噪比。通过提取能较好反映雷达目标散射点回波特性的多维特征向量，设计BP神经网络作为分类器，提出了一种基于目标多维特征向量以及BP神经网络的高分辨雷达目标识别方法。利用在微波暗室测量获得的三种国产飞机模型回波数据进行目标识别处理，实验结果表明，提出的方法能有效地完成三种目标识别任务，在虚警率低于3%的情况下正确识别率优于95%。

关键词: 目标识别, 高分辨距离像, 序贯预处理, 反向传播（BP）神经网络, 多维特征向量