|
|||||
|
|
| RBF神经网络在遥感影像分类中的应用研究 | |
| 引用本文： | 罗小波,王云安,肖春宝,王西林. RBF神经网络在遥感影像分类中的应用研究[J]. 遥感技术与应用, 2004, 19(2): 119-123 |
| 作者姓名： | 罗小波 王云安 肖春宝 王西林 |
| 作者单位： | 1. 中国地质大学信息工程学院,湖北,武汉,430074 2. 湖北省交通规划设计院,湖北,武汉,430050 3. 中国石油天然气股份有限公司勘探开发研究院西北分院,甘肃,兰州,730020 |
| 摘 要： | 用RBF神经网络进行遥感影像分类，在网络结构设计上使RBF层与输出层的节点数都等于所要分类的类别数。用Kohonen聚类算法确定RBF中心的时候，用训练样本的均值作为初始中心，并在RBF宽度进行求取的时候进行了改进，以避免内存溢出。所设计的RBF神经网络分类模型具有结构简单、算法简洁的优点。实验结果表明，该方法用于遥感影像分类取得了较高的分类精度，具有实际应用价值。 |
| 关 键 词： | 遥感影像 RBF神经网络 监督分类 |
| 文章编号： | 1004-0323(2004)02-0119-05 |
| 修稿时间： | 2003-07-21 |
Application Study on Classification of Remote Sensing Imagery Using RBF Neural Network |
|
| LUO Xiao-bo,WANG Yun-an,XIAO Chun-bao,WANG Xi-lin. Application Study on Classification of Remote Sensing Imagery Using RBF Neural Network[J]. Remote Sensing Technology and Application, 2004, 19(2): 119-123 | |
| Authors: | LUO Xiao-bo WANG Yun-an XIAO Chun-bao WANG Xi-lin |
| Affiliation: | LUO Xiao-bo~1,WANG Yun-an~2,XIAO Chun-bao~1,WANG Xi-lin~3 |
| Abstract: | Compared with BP neural network , RBF neural network has some advantages, such as fast training speed and not being easy to plunge into local minimum.In the same time, it is not easy to set the number of RBF layers' nodes. In this paper, we have made some improvement on RBF neural network when using it in classification of remote sensing imagery. During the structure of network being designed, the number of RBF layers' nodes and the number of output layers' nodes are both equal to the number of classes to be classified. Doing so, the difficulty to set the number of RBF layers' nodes is no longer exsting. And for a given practical problem,the structure of RBF neural network is determinate. In Kohonen algorithm, the initial weights is randomly set. But other experiments show that the iniatial weights is important to the training of RBF neural network.In this paper, when using Kohonen algorithm to train RBF center, we use the mean values of training samples as the initial center of RBF, and make some mending to avoid memory overflow when computing the width of RBF. Experimental result shows that the accuracy of classification of this classification model is comparatively high, and it has practical application value. |
| Keywords: | Remote sensing imagery RBF neural network Supervised classification |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ | |
