摘要
提出了一种基于改进BP神经网络的混合像元分解模型,通过加入动量项和变步长法对标准BP算法进行改进。利用该模型对多光谱遥感影像进行了处理分析,与一般BP分解模型进行比较后证明:改进的BP神经网络模型具有正确率高,自适应能力强的优点。
It is difficult to classify mixed cells in remote sensed image. In this paper, a modified BP neural network was developed to resolve it through adding a momentum and changing the learning rate. This algorithm was tested on a muhi-spectral image. Obtained accuracy measures were higher compared with conventional BP un-mixing model, and it has been proved effective.
出处
《地理空间信息》
2009年第3期65-67,共3页
Geospatial Information
