摘要
由于在不同的观察角度、位置以及光照等条件下雷达目标图像之间差异较大,使得很多经典的维数约简和特征提取算法不能有效地用于飞机目标图像识别。基于二维局部敏感判别分析(2DLSDA),提出了一种雷达目标识别方法。首先构造类内和类间邻域关系图,计算两个邻域图上的权重矩阵;然后基于Schur分解求出两个正交变换矩阵,得到映射矩阵,对观察数据进行维数约简,由此有效地克服小样本问题。对飞机目标的分类实验结果表明,该方法是有效可行的。
Since the images of an aircraft target are much different from each other under various conditions of different observed angle, locality and illumination, many classical dimensional reduction and feature extracting methods are not effective to recognize the aircraft target. A recognition method of radar target is proposed based on two-dimensional locality sensitive discriminant analysis (2DLSDA). Firstly, two graphs respectively representing intra-class and inter-class neighbor relationship are constructed. Then, weight matrixes are calculated out. Finally, two orthogonal transform matrixes are computed out based on Schur decomposition. The projection matrix is obtained and then the dimensionality of the image is reduced. Thus the small-sample-size problem can be overcome. The recognition results on radar targets show that the proposed method is very effective and feasible.
出处
《电光与控制》
北大核心
2013年第4期10-12,共3页
Electronics Optics & Control
基金
河南省教育厅科学技术研究重点项目(12B120012)
关键词
雷达
目标识别
二维局部敏感判别分析
维数约简
radar
object recognition
2D locality sensitive discriminant analysis (2DLSDA)
dimensionality reduction