摘要
首先对空间目标辐射特性进行了研究,指出了用空间目标在3个不同波段的辐射通量作为特征向量进行目标识别。然后研究了进化模糊神经网络(EFuNN)和动态进化神经模糊推理系统(DENFIS),最后用EFuNN和DENFIS进行了仿真实验,并且与BP神经网络、遗传算法以及遗传-神经算法进行了比较。仿真结果表明EFuNN尤其是DENFIS具有较好的学习能力和泛化能力,较大地提高了目标识别率,能够较好地进行空间点目标的识别。
Radiometric characteristics for spacial target are discussed.And irradiances at three different wave bands of spacial target are taken as eigenvector for target recognition.Tben,EFuNN and DENFIS are presented for spaciai point target recognition. At last numerical experiment based on EFuNN and DENFIS is presented.It is demonstrated that EFuNN,especially DENFIS can effectively learn complex temporal sequences in an adaptive way and outperform some existing models.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第32期244-248,共5页
Computer Engineering and Applications
关键词
模糊神经网络
进化模糊神经网络
动态进化神经模糊推理系统
Fuzzy Neural Networks （FNN）
Evolving Fuzzy Neural Network （EFuNN）
Dynamic Evolving Neural-Fuzzy Inference System（DENFIS ）
