基于改进小波变换的地物类型提取研究

Study on the Extraction of Land Coverage Type Based on Improved Wavelet Transform

在多源遥感图像融合过程中,为充分利用高分辨率影像的低频信息,从而提高对影像的解译能力,本研究尝试对传统的小波融合算法进行了改进,以SPOT5的多光谱波段和全色波段为数据源,借助MatLab与ENVI等工具,实现改进后小波变换的遥感影像融合,该融合方法跟传统融合方法相比,信息熵和清晰度都有所提高。分别以传统融合图像及改进小波变换融合后图像为分类底图,实施了监督分类,完成地物类型的提取,结果表明:改进的小波变换融合后图像分类效果更好,分类精度和Kappa系数都有所提高,其中利用小波变换融合为基础的马氏距离法和决策树算法相结合的分类效果最佳。

In order to take advantage of low frequency information of high-resolution image and improve the ability of image interpretation during fusing the multi-source image,in this study,taking multi-spectral bands and panchromatic band of SPOT5 as the data source,using MatLab and remote sensing software ENVI as analysis tools,the traditional wavelet fusion algorithm is improved and is used for the image fusion at the same time,comparing to the traditional fusion methods,the entropy and clarity of improved wavelet fused image are improved distinctly,the extraction of land coverage type are accomplished according to the supervised classification using the improved wavelet fused image and traditional fused image,the results show that the classification effects which use the improved wavelet fusion image are better,the whole classification accuracy and kappa coefficient are improved among these methods.The effect of combing methods of Mahalanobis Distance and Decision Tree basing on improved wavelet fused image is the best.