基于多时相遥感影像的土地利用变化检测研究

发布时间：2018-02-12 01:15

本文关键词： 变化检测土地利用对象特征特征融合数据维压缩监督分类　出处：《昆明理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：及时、快速、准确地获取土地利用信息是各级政府开展国土资源规划、管理、保护和合理利用的前提条件和基本依据。利用多源,多时相,多分辨率遥感影像进行土地利用变化检测,可获取多目标、多周期、多尺度的变化信息。论文以同一个研究区域不同范围的QuickBird影像和Landsat多光谱数据为研究对象,从两个方面展开研究：一、在深入研究常用遥感变化检测方法的基础上,针对传统变化检测方法在高分辨率遥感影像变化检测中的局限性,利用面向对象分析技术对实验区多时相QuickBird影像进行变化检测实验,检测出变化图斑后与实验区更高分辨率的航摄影像及历史矢量数据进行对比,来分析该方法在土地利用变化检测中的适用性；二、基于需要了解某个时间段研究区域内的土地利用变化趋势但又缺乏历史矢量数据及高分辨率遥感影像的假设,以Landsat多光谱数据为研究对象,结合相应的指数,利用压缩数据维的方法进行分类后变化检测,获取研究区域主要土地利用类型的变化信息,其具体研究内容及成果如下： (1)对遥感变化检测的基本理论进行深入研究,并概括分析了常用变化检测方法的原理及优缺点。 (2)以研究区域某个村为实验对象,对实验区多时相QuickBird影像进行多尺度分割后,提取影像对象的光谱特征、纹理特征和形状特征图,利用提取的对象特征进行变化检测,由于单一的特征所提取的信息具有一定的不确定性,不能唯一地准确判定地表的变化信息,实验将特征级变化检测结果进行融合作为最终的检测结果。后将实验结果与实验区0.2米航片、土地利用变更调查及基本农田规划数据制作的本底数据进行对比分析,结果表明此方法能最大程度的检测变化信息,但要运用到实际的土地利用变化检测项目中,需要对变化检测结果进行大量的筛选。 (3)以整个研究区域为实验对象,通过分析研究区域主要的土地利用类型和地物分布特点,选取土壤调节植被指数、归一化建筑用地指数及修正归一化差异水体指数代表实验区的三大主要土地利用类型—植被、建筑用地及水体。将TM影像6个波段及ETM影像7个波段压缩为由它们衍生的3个采用比值运算构成的指数波段,并分析三个指数波段的相关性,将相关性很小的两个波段作为XY轴构成二维散点图,利用基于几何顶点的端元提取方法选择分类样本后对指数影像进行分类后变化检测。实验结果表明基于相同的分类样本和分类方法,压缩后的指数影像分类精度高于原始影像,利用压缩数据维的分类后变化检测方法可有效检测研究区域主要土地利用类型的变化信息。
[Abstract]:Timely, rapid and accurate access to land use information is the prerequisite and basic basis for governments at all levels to carry out land and resource planning, management, protection and rational utilization. Multi-resolution remote sensing images can obtain multi-target, multi-period and multi-scale information for land use change detection. In this paper, QuickBird images and Landsat multispectral data of different range in the same research area are taken as the research object. The research is carried out from two aspects:. First, on the basis of in-depth study of common remote sensing change detection methods, aiming at the limitation of traditional change detection methods in high-resolution remote sensing image change detection, The object oriented analysis (OOA) technique is used to detect the change of multitemporal QuickBird images in the experimental area, and the change pattern is detected and compared with the higher resolution aerial photography and historical vector data in the experimental area. To analyze the applicability of this method in land use change detection; Secondly, based on the assumption that land use change trends in a certain period of time need to be understood but lack of historical vector data and high-resolution remote sensing images, the multi-spectral data of Landsat are taken as the research object and the corresponding indices are combined. The method of compressed data dimension is used to detect the changes after classification and obtain the change information of the main land use types in the study area. The specific research contents and results are as follows:. 1) the basic theory of remote sensing change detection is deeply studied, and the principle, advantages and disadvantages of the commonly used change detection methods are summarized. Taking a village in the research area as the experimental object, the multitemporal QuickBird image of the experimental area is segmented with multiple scales, then the spectral feature, texture feature and shape feature map of the image object are extracted, and the change detection is carried out by using the extracted object feature. Because the information extracted by a single feature has certain uncertainty, it is not possible to judge the change information of the earth's surface only accurately. In the experiment, the feature level change detection results are fused as the final detection results, and then the experimental results are compared with the background data of 0.2 m aerial photograph, land use change survey and basic farmland planning data in the experimental area. The results show that this method can detect the change information to the maximum extent, but in order to apply to the actual land use change detection project, a large number of change detection results need to be screened. Taking the whole study area as the experimental object, through analyzing the main land use types and the distribution characteristics of the ground objects in the study area, selecting the soil adjustment vegetation index, Normalized construction land index and modified normalized difference water body index represent the three main land use types in the experimental area-vegetation. Six bands of TM image and seven bands of ETM image are compressed into three exponential bands derived from TM image by ratio operation, and the correlation of the three exponential bands is analyzed. Using the two bands with little correlation as the XY axis to form a two-dimensional scatter diagram, Based on the geometric vertex extraction method, the classification samples are selected and the changes of the index images are detected after classification. The experimental results show that the classification accuracy of the compressed exponential images is higher than that of the original images based on the same classification samples and classification methods. The change information of the main land use types in the studied area can be detected effectively by using the classified change detection method of compressed data dimension.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P237

【引证文献】