摘要
针对雷达目标一维距离像的平移敏感性和姿态敏感性,提出一种提取一维距离像的偶数阶中心矩作为雷达目标特征的方法。用小波变换提高一维距离像的信噪比,在此基础上提取中心矩特征,再选取维数减半的比较稳定的偶数阶中心矩作为目标特征,以适用于支持向量机分类器进行识别分类。对实测雷达目标的数据进行试验,结果显示在减少模板特征向量的存储量和测试样本识别时的计算量的同时,得到了较高的识别率。
A computationally efficient method is proposed for radar target recognition based on even rank central moments of high resolution range profile. Firstly, wavelet denoising is used to enhance the signal noise rate of high resolution range profile. Then, translation-invar- iant central moments are extracted from the denoised range profile. And then even rank cen- tral moments, with dimension-halved and relatively stable property, are selected as target features. A multi-class support vector machine classifier is designed to classify radar targets. The experimental results based on real radar data show that the proposed method achieves high recognition rate, which reduces the storage of the template feature vectors and calcula- tion of the test sample identification.
出处
《沈阳理工大学学报》
CAS
2013年第5期38-42,47,共6页
Journal of Shenyang Ligong University
关键词
雷达目标
一维距离像
偶数阶中心矩
支持向量机
radar target
high resolution range profile
even rank central moments
support vector machine