论文部分内容阅读
反向传播神经网络（ＢＰ网络）能解决传统统计分类方法的不足，现已逐渐用于遥感图像分类中，研究用一种新的改进ＢＰ算法进行遥感图像分类。方法用线性搜索的共轭梯度法（ＣＧＬ）动态选取学习速率以提高训练速度，结果计算机仿真表明，在分类精度未下降的情况下，训练时间较其它改进算法减少５一１１０ｓ．结论该方法避免了存储量大的负担及误差函数的发散，适用于遥感图像的分类。
Backpropagation neural network (BP network) can solve the shortcomings of traditional statistical classification methods, and has been gradually used in remote sensing image classification. A new improved BP algorithm for remote sensing image classification is studied. Methods Using the linear search of conjugate gradient method (CGL), the learning rate was dynamically selected to improve the training speed. The results of computer simulation showed that the training time was reduced by 5 to 110 s compared with other improved algorithms when the classification accuracy did not decrease. Conclusion This method avoids the burden of large amount of storage and the divergence of error function and is suitable for the classification of remote sensing images.