长汀县Spot5影像多尺度分割与信息提取方法研究

发布时间：2018-01-15 02:08

本文关键词：长汀县Spot5影像多尺度分割与信息提取方法研究　出处：《福建师范大学》2015年硕士论文　论文类型：学位论文

【摘要】：本文以Spot5高分辨率卫星遥感影像作为数据源,以福建省长汀县为试验区,开展了对长汀县2010年Spot5高分辨率遥感影像的多尺度分割和地物信息提取的研究。首先,针对Spot5卫星遥感影像的特点,进行遥感影像的预处理。然后对影像进行多尺度分割,确定最合适的分割尺度和形状因子、紧凑度因子设置值,得到适宜各地物信息提取对象,并根据不同分割尺度确定各地物类别的分类层次。最后,结合使用模糊数学分类方法和最邻近分类方法得到分类结果,并使用混淆矩阵对分类结果进行精度分析。得到如下结论：(1)多尺度分割参数设置上,“尺度”为150时对河流、道路和工厂厂房能够呈现比较完整且均质的对象；“尺度”为50时建筑物、耕地和植被覆盖地等呈现比较理想的分割对象。(2)分割参数试验中当形状因子设置为0.2,紧凑度因子设置为0.5时,分割的影像对象最为理想。(3)植被覆盖地信息提取模糊规则为“标准化植被指数”=0.06且“亮度”89,水体信息提取模糊规则为65“近红外波段的光谱均值3”78.3,道路信息提取模糊规则为0.3“密度”1,农田信息提取模糊规则为-0.006“标准化植被指数”0.09,建筑用地信息提取规则为1“边境指数”1.82且0.03“纹理同质性”0.13,对剩余未能很好提取的类别选取训练样本进行最邻近分类法进行提取。(4)最后将提取结果以长汀县2011年国土调查数据为参考进行精度评价,得到总体分类精度93.73%,Kappa系数为0.77。
[Abstract]:In this paper, Spot5 high-resolution satellite remote sensing image is used as the data source and Changting County in Fujian Province as the experimental area. In this paper, the multi-scale segmentation and feature extraction of Spot5 high-resolution remote sensing images in Changting County in 2010 were studied. Firstly, the characteristics of Spot5 satellite remote sensing images were analyzed. The preprocessing of remote sensing image is carried out. Then the multi-scale segmentation is carried out to determine the most appropriate segmentation scale and shape factor and the set value of compactness factor. And according to the different segmentation scale to determine the classification level of the local categories. Finally, combined with the use of fuzzy mathematics classification method and the nearest neighbor classification method to obtain the classification results. The accuracy of the classification results is analyzed by using the confusion matrix. The following conclusion is drawn: 1) when the scale is 150, the multi-scale segmentation parameters are set up for the rivers. Roads and factory buildings can present relatively complete and homogeneous objects; When "scale" is 50, buildings, cultivated land and vegetation cover show ideal segmentation object.) in the experiment, when the shape factor is set to 0.2, the compactness factor is set to 0.5. The fuzzy rule of extracting vegetation cover information is "standardized vegetation index" (0.06) and "brightness" (89). The fuzzy rule of extracting water information was 65 "near infrared spectral mean 3" 78.3, and the fuzzy rule of road information extraction was 0.3 "density" 1. The fuzzy rules of farmland information extraction were -0.006 "standardized vegetation index" 0.09, and the rules of information extraction of construction land were 1 "border index" 1.82 and 0.03 "texture homogeneity" 0.13. At last, the precision evaluation was carried out based on the land survey data of Changting County in 2011. The overall classification accuracy of 93.73 kappa coefficient is 0.77.
【学位授予单位】：福建师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P237

【参考文献】