BP神经网络的沙漠化土地信息提取研究

以塔克拉玛干沙漠南缘策勒绿洲为例,探讨了基于主成分融合的沙漠化信息的提取方法.由于Landsat-7 ETM 的全色波段与多光谱波段有相同成像条件,影像获取时间一致,两种不同分辨率的数据可以不经配准而实现高精度融合.首先,对Landsat-7ETM 的全色图像与多光谱图像进行主成分融合处理,再利用BP神经网络模型,以相同的训练样本分别对融合前后的影像进行分类,在此基础上进行沙漠化信息的提取.结果表明:主成分变换融合图像的光谱信息保持性、信息量以及空间分解力都较高,且分类精度比Landsat-7ETM 多光谱图像有较大提高,是监测沙漠化土地变化的有效手段.

Study on Extracting the Information about Desertified Lands Based on Principal Component Fusion and BP Neural Network

Sandy desertification is a severe environmental problem in arid areas,which threatens certainly the stability and sustainable development of oases,so it is very important to timely wxtract the information and spatial distribution of sandy desertification by applying remote sensing means.In this paper,the PCA data fusion methods and BP network classification for deriving the information of desertified lands in Qira Oasis located in the central southern marginal zone of the Taklimakan are researched.Because the Landsat-7 ETM Panchromatic band and the multispectral bands can be simultaneously obtained from the same sensor system under the same imaging conditions,it is available to get best data fusion results without registration.Firstly,Principal Component Analysis(PCA) is used to fuse the panchromatic and multispectral images of Landsat-7 ETM Satellite.And then,the BP Neural Network classification method is applied to classify the Landsat-7 ETM multispectral bands and the fusion images with the same training samples.The results reveal that PCA fusion image has more information and higher spatial resolution while maintaining the basic spectral characteristics of the original image,and also has a higher classification accuracy of the results than the Landsat-7 ETM multispectral image.It is useful to detect and monitor sandy desertification change in arid areas.