基于流形学习的有监督雷达目标成像识别算法  

One-DimensionalityRadar Target ImagingRecognition Supervision Altorithm Based on Manifold Learning

作　　者：耿伯英[1] 崔萌[1] 梁洁[1] 

机构地区：[1]海军工程大学,武汉430033

出　　处：《计算机与数字工程》2016年第12期2404-2406,2436,共4页Computer & Digital Engineering

摘　　要：运用经典流形学习算法ISOMAP对高分辨率雷达一维像进行非线性的降维预处理,通过广义回归神经网络学习逼近流形降维的映射关系,利用概率神经网络完成目标识别。仿真实验证明该算法对具有非线性展布形式的数据集有较好的识别效果。The classical manifold learning algorithm ISOMAP is used to do the nonlinear dimension reduction pretrea-ment to the high resolution radar pictures. The mapped function between the high-dimensionality manifold and the lower one is drawn up by generalized regression neural network, and the recognition is completed by probabilistic neural network. Em- ulational experiments proves the good recognition effect to the data sets which has nonlinear distribution

关 键 词：流行学习 ISOMAP 广义回归神经网络 概率神经网络 

分 类 号：TN957[电子电信—信号与信息处理]