摘要
一维距离像是自动目标识别的一种重要特征,它对目标姿态变化很敏感,只有通过进一步处理提取稳定特征才能够有效用于识别。针对距离像的这种姿态敏感性,首先分析了主分量分析(PCA)的降噪原理与核主分量分析(KPCA)的特征提取能力,然后提出先用PCA滤波对一维距离像降噪再用KPCA提取较大姿态角范围内稳定特征的雷达目标一维距离像识别框架,并用四类目标的实测数据进行分类实验,表明该算法确实能够提高识别性能。
One-dimensional range profile is an important feature for automatic target recognition, which is very sensitive to attitude changes of the target. Only by further processing and feature extracting can it be used for effective target recognition. In this paper, the principles of Primary Component Analysis （PCA） based noise-reduction and the powers of Kernel Primary Component Analysis （KPCA） for feature extraction are analyzed in detail. Then, a framework for automatic radar target recognition based on one-dimensional range profile is put forward, in which PCA filtering is used for noise-reduction of the range profile, and KPCA is proposed for extracting the stabilization feature over a large attitude angle. The results of recognition experiment with targets of four types show that the approach can really improve the performance of recognition.
出处
《电光与控制》
北大核心
2005年第5期28-31,共4页
Electronics Optics & Control
作者简介
张仲明（1980-），男，浙江浦江人，硕士生，主要研究方向为雷达目标的特征提取与自动识别。
