基于TM图像的城市不透水面信息提取

摘要

Based on analyzing the theory of the Optimum Band Combination, Principal Component Analysis (PCA) and NDISI, this paper presents an improved method, i. e. , "experimental layer stack", to extract impervious surface of Taiyuan city, Shanxi Province, from Landsat TM image. Both unsupervised and supervised classification methods were used to classify the original multi - band image, PCA image, NDISI and experimental band combination images. The accuracies of the classification were assessed using 256 sampling points randomly selected from Google Earth high resolution image of Taiyuan. By comparison and analysis, the authors found that the experimental B combination method obtained the highest overall accuracy of 87.72% with the Kappa coefficient of 0.85.%基于TM图像,利用遥感技术,以山西省太原市为研究区域,提取不透水面信息.综合比较和分析了被广泛应用的主成分分析法和归一化差值不透水面指数法,进而提出了一种改进的不透水面提取方法——实验图层组合法.利用随机生成的256个分类评价采样点,以可提供经纬度信息的Google Earth作为参考,将通过监督分类方法获得的分别基于原始多波段图像以及主成分分析法、归一化差值不透水面指数法和实验图层组合法得到的分类结果进行了评估,获得了4种图像的分类精度.比较可知,实验图层组合法总体分类精度高于其他3种结果,为87.72％,Kappa系数为0.85.