【作者】 顾晓文；

【导师】 王建弟；

【作者基本信息】 浙江大学 ， 土壤学， 2004， 硕士

【摘要】 土地资源是人类赖以生存和发展的基础，是国土资源的重要组成部分。20世纪以来，人口增长，经济发展，人地矛盾日益突出，人们对土地的不合理利用导致土地退化加剧，全球环境恶化，社会经济的可持续发展受到严重影响。因此，土地资源的利用及其变化动态受到普遍的关注。 人口多，耕地少，耕地后备不足是我国的基本国情。加强土地管理，切实保护耕地是我国的治国之策。全国范围的土地利用现状调查已基本结束，各地今后的工作主要是土地变更调查与统计。土地利用变更调查是土地管理中最基础的最重要的环节，是一个极其复杂的工作，需要处理大量数据，如何提高对土地信息更新特别是用途最大的1：1万土地利用信息的效率和质量，对加快土地管理信息化建设具有重要意义。 遥感技术以其宏观性、实时性、周期性及综合性等特点，为快速、客观、准确地更新土地利用现状信息提供了可能。但是由于遥感信息的综合性、遥感成像机理的复杂性，信息处理技术远落后于获取技术的发展，自动化、高精度、智能化土地利用信息解译仍未实现，自动解译的精度往往低于目视解译的精度。在从1999年开始的新一轮国土资源大调查中，高分辨遥感资料成为土地利用信息更新的主要数据源，到2002年，分辨率10m的SPOT5卫星多光谱数据和2.5m全色数据已成为2002年土地利用动态监测的主要数据源。如何最有效、充分利用SPOT5影像来更新土地利用信息，目前在应用研究上研究还比较少。 本文以浙江省台州市椒江区为研究对象，利用SPOT5 10m的多光谱数据和2.5m全色数据，针对土地利用遥感分类中的影像处理，多源影像融合、影像分类方法和土地利用现状信息如何更新等问题开展研究，以期获得适合利用SPOT5卫星影像来更新1：1万土地利用现状信息的技术方法。 研究的主要内容和结论如下： 1．研究中充分使用了遥感信息与非遥感信息的结合来进行土地利用现状信息中文摘要的更新。将纸质的地形图经扫描、几何校正及色彩归化后，形成在内容、几何精度和色彩等方面与地形图基本保持一致的栅格数据文件。在此基础上获得椒江的数字高程模型，并将它们统一在相同的坐标体系中，为后面的几何校正创造了基础。在监督分类的时候，将土地利用现状矢量图与欲分类影像叠置进行训练样本的选取以及抽样点的选取。 2.在试验中根据数据源及数据的可靠性，决定选择最适合SPOTS影像的校正方法。椒江区虽以海积平原为主，但是中间间布小山与陆屿，采用多项式纠正后影像与土地利用现状矢量叠加的效果较差，因此在几何校正的时候采用了基于共线方程的正射校正方法，用ERDAS的OrthoBASE模型完成。 3.在GIS数据支持下，对SPOTS的1 om原始多光谱影像和融合后重采样的2.5m多光谱影像分别进行监督分类试验，精度分析表明，融合后重采样的多光谱影像的分类总体精度只能达到73.80%，远低于原始多光谱影像的分类总体精度89.20%，不适合将分类结果继续运用在对1:1万土地利用现状信息的更新工作中。监督分类精度结果表明，利用SPOTS高分辨率融合影像进行土地利用现状信息更新，不能使用将分类栅格图像直接经拓扑关系转换成精度较高的矢量图像的方法，而人机交互目视解译是一个最佳方法。更多还原

【Abstract】 Land is the most basic resource for our living and development and it is also an important part of our country resources. Since the beginning of 20th century, the world population has dramatically increased. With the rapid economical development, the lack of land is becoming a bottleneck for our further development. At the meantime, a lot of land resources are not optimally used. This leads the decrease of the land quality and the deterioration of the world environment. The continual development of the economy has been seriously blocked. Because of its importance, how to reach a good land resource use and how to monitor its change get more and more attention from different fields.In our country, we are facing the fact that there is a great amount of population and we do not have enough plantations. Therefore we need a good land resource management to achieve an optimal use with the current plantations. A countrywide investigation of the current land use has been finished. The future emphasis will be the investigation and statistics of the change of land use. This work is the most basic and important in the lifecycle of the land management. It is a complex work and needs a great amount of actual data. With the information retrieved in the investigation phase, we can speed up the information evolution in our land management.Remote sensing technology has features such as wide-region, real time monitoring, periodical information, and comprehensive information. All these features give the possibility for a fast, objective, precise updating of land-use. Because of the wide scope of remote sensing information and the complexity of remote sensing image generation process, the remote sensing information processing technology is far behind the remote sensing technology itself. An automatic, precise, and intelligent processing technology for land use information retrieved by using remote sensing technology is still not mature. In most cases, the precision of this process is lower than the precision done by human eyes. Since the new land use investigation in 1999, high-resolution image data has become the most important data resource of land use updating. Until 2002,10m-resolution multispectral image from SPOT5 satellite and 2.5m PAN data have been the main data resource for dynamic monitoring of land use. But currentlythere are not so many researches being done to find out how to effectively use SPOT5 satellite images to update the land use information.This thesis takes the Jiaojiang region of Taizhou in Zhejiang province as an example. By using SPOT5 10m multispectral image data and 2.5m PAN data, a series studies have been done on how to rectify the remote sensing image data, how to merge images from different resources, how to classify the different image data, and how to update the land usage information. The goal of the studies is to get a feasible technology to update 1:10000 land use information by using SPOT5 satellite images.The carried out studies and the conclusions are described below:1. A study has been carried out to find out how to combine the remote sensing information and the non-remote sensing information to update the land use information. The paper topographical map is firstly to be scanned geometric corrected. Next a DRG with the same content, geometric precision and color as topographical map is created. On the base of this DRG, a DEM of Jiaojiang is generated. DRG and DEM are put in the same coordination system; this provides the basis for the later geometry rectification. During the supervised classification, the selection of samples is also applied on land use vector information.2. On the basis of reliability of data resources, the most proper method to rectify the SPOT5 image is selected. Jiaojiang region has mainly reformed by sea alluviation, but it also has some mountains here and there. The result of folding image on land use information after polynomial correction is not as expected. Therefore, a correction method based on the collinear equation is applied in the model of OrthoBA更多还原