基于神经网络的舰船雷达目标特征抽取与分类方法研究

The Methods for Feature Extraction and Classification of Radar Ship Target Using Neural Network

本文着重研究应用神经网络来进行舰船雷达目标特征抽取与分类问题,提出了一种基于Mellin变换和多层前馈神经网络的特征抽取方法和一种基于Kohonen网络组的特征分类方法。采用实地录取的三类舰船雷达目标视频回波数据对本文提出的有关方法进行检验,结果表明本文提出的雷达目标特征抽取与分类的神经网络方法是切实可行的,其抽取的特征具有良好的稳定性,其分类的精度很高,明显优于传统的K-邻近分类器。

The problem of applying neural network to the feature extraction and classification of radar ship target is discussed in this paper. A new feature extraction method is proposed based on Mellin transform and multi-layered feedforward neural network. A new type of neural network classifier is designed due to a group of Kohonen networks. Experiments are carried out for the proposed methods with practical incoherent radar ship target video-echo data and the corresponding results indicate that: a)The features extracted by the proposed method have both scale- and shift-invariance, and b) The classifier designed in the paper has much higher classification accuracy than K-nearest neighbour classifier.