摘要
针对超谱图像高维光谱信息给传统分类带来的困难,结合径向基神经网络的原理,提出了一种概率神经网络分类方法。并将其成功应用到具体超谱图像数据中,验证了概率神经网络分类器的有效性。通过实验仿真,研究了特征向量维数对分类结果的影响,证明概率神经网络可应用于大于100个波段的超谱图像数据。
Aiming at based on the theory of method is illustrated, tion. research on the the difficulty of traditional classification brought by hyperspectral image's high dimension, radical basis function network, a kind of PNN (Probability Neural Network) classification and its affectivity is verified by its application in hyperspectral image. Through the emula-effect of feature image's dimensions to classification result, proves the PNN's applicability high dimension hyperspectral image.
出处
《吉林大学学报(信息科学版)》
CAS
2008年第2期122-125,共4页
Journal of Jilin University(Information Science Edition)
基金
国家自然科学基金资助项目(60302019)
关键词
概率神经网络
超谱图像分类
特征矢晕维数
probability neural network (PNN)
hyperspetral image classification
feature image's dimensions