数据融合方法在城市遥感监测中的应用研究

发布时间：2018-06-21 22:19

本文选题：数据融合 + 不透水面　；参考：《内蒙古师范大学》2017年硕士论文

【摘要】：数据融合是将遥感影像经过融合处理,以提高研究区影像的空间以及时间分辨率。将数据融合方法应用于城市遥感监测,可以提高城市信息提取的精度,对快速获取城市信息具有重要意义。本文利用Landsat 8 OLI-TIRS影像数据作为数据源,提取研究区内城市信息,以进行不同数据融合方法在城市遥感监测中的应用评价研究。本文将内蒙古呼和浩特市市区南部与农村交界的小面积矩形区域作为研究区(111°36′47″E—111°44′08″E,40°45′33″N—40°48′12″N),经过遥感影像数据预处理后,分别进行HSV变换、Brovey变换、Gram-Schmidt变换、主成分变换、NNDiffuse Pan Sharpening变换等5种数据融合处理,然后对5种融合后影像进行基于支持向量机的城市遥感信息提取,并对Gram-Schmidt变换、主成分变换、NNDiffuse Pan Sharpening变换3种融合后的影像进行基于归一化差值不透水面指数的城市遥感信息提取。通过对融合后的影像进行参数统计分析(包括均值、标准差、信息熵、平均梯度、相关系数),以及对基于融合影像的信息提取结果(包括基于支持向量机的城市信息提取结果和基于归一化差值不透水面指数的城市信息提取结果)进行精度分析,确定最适宜城市信息提取的数据融合与信息提取方法。研究结果表明,在综合考虑融合后影像的各个统计参数的情况下,NNDiffuse Pan Sharpening融合的效果最好,其影像在各个波段的保真度、清晰度以及信息丰富程度都较其它融合方法高;对基于支持向量机的城市信息提取结果进行精度分析,发现基于NNDiffuse Pan Sharpening的支持向量机分类结果精度最高,其次是基于PC的支持向量机分类结果精度,最差的是基于HSV和Brovey的支持向量机分类结果精度;在基于归一化差值不透水面指数的城市信息提取结果中,总体上基于PC变换的NDISI提取结果精度最高,其次是基于NNDiffuse Pan Sharpening变换的NDISI提取结果,最差的是基于Gram-Schmidt变换的NDISI提取结果;对所有信息提取方法进行精度评价,发现基于PC变换的NDISI提取结果的精度最高。
[Abstract]:Data fusion is the fusion of remote sensing images to improve the spatial and temporal resolution of the image. The application of data fusion method to urban remote sensing monitoring can improve the precision of urban information extraction, and it is of great significance to obtain urban information quickly. In this paper, Landsat 8 OLI-TIRS image data is used as the data source to extract urban information in the study area and to evaluate the application of different data fusion methods in urban remote sensing monitoring. In this paper, the small rectangular area at the southern and rural border of Hohhot, Inner Mongolia, is taken as the study area (111 掳36N 47 "E-111 掳44008" EN 40 掳4510 33 "N-40 掳4812" N), and the data of remote sensing image are preprocessed. Five kinds of data fusion processing, such as HSV transform, Brovey transform, principal component transform and NNDiffuse Pan Sharpening transform, are carried out respectively. Then, five kinds of fused images are extracted from urban remote sensing information based on support vector machine, and Gram-Schmidt transform is used. Principal component transform (PCA) and NNDiffuse Pan Sharpening transform are used to extract urban remote sensing information based on normalized difference impervious surface index. Through the statistical analysis of the fused image parameters (including mean, standard deviation, information entropy, average gradient), Correlation coefficient), and the accuracy analysis of the information extraction results based on fusion image (including the results based on support vector machine (SVM) and city information extraction based on normalized difference impervious surface index (NDI). The method of data fusion and information extraction which is most suitable for urban information extraction is determined. The results show that the NNDiffuse Pan Sharpening fusion has the best effect when considering all the statistical parameters of the fused images, and the fidelity, clarity and information richness of the images in each band are higher than those of other fusion methods. By analyzing the accuracy of urban information extraction based on support vector machine (SVM), it is found that the accuracy of SVM classification based on NNDiffuse Pan Sharpening is the highest, and that of SVM based on PC is the highest. The worst one is the classification accuracy of support vector machine based on HSV and Brovey, and the NDISI based on PC transformation has the highest accuracy in the city information extraction based on normalized difference impervious surface index. Secondly, NDISI extraction results based on NNDiffuse Pan Sharpening transform, the worst NDISI extraction results based on Gram-Schmidt transform, and the accuracy evaluation of all information extraction methods show that the accuracy of NDISI extraction based on PC transform is the highest.
【学位授予单位】：内蒙古师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】