基于BP神经网络的多源遥感影像分类

摘要

在研究人工神经网络理论的基础上，应用动量法和学习率自适应调整的策略，改进了BP神经网络法。并用于对同一地区的Landsat TM3,4,5影像和航空SAR影像融合进行分类和分类融合结果进行了比较。结果表明：同标准的BP神经网络、传统的Bayes融合分类法相比，改进的BP神经网络融合法不仅获得了标准BP网络高的分类精度，可同Bayes融合媲美，而且提高了学习率，增强了算法的可靠性，因而提高了影像分类速度，更适用于遥感影像分类。%A fusion method for target recognition based on artificial neurual B-P network are studied. This improved method is used in classification of land use with remote sensing imagery such as SAR and TM band 5,4,3. Compared with classification of the standard B-P network and Bayesian statistics, the results show it has not noly the highest accurracy but also the fastest speed of classification. So it is applied in classification of remotely sensed images.