基于时序遥感影像和面向对象分类技术的城市扩展信息提取研究
文献类型：学位论文
| 作者 | 兴光 |
| 学位类别 | 硕士 |
| 答辩日期 | 2008 |
| 授予单位 | 中国科学院水利部成都山地灾害与环境研究所 |
| 授予地点 | 成都 |
| 导师 | 李发斌 |
| 关键词 | 时序遥感影像 面向对象分类技术 城市扩展信息提取 |
| 其他题名 | Study on Extracting Urban Expansion Information from Sequential RS Data Based on Object-oriented Technology |
| 学位专业 | 地图学与地理信息系统 |
| 中文摘要 | 城市是人口、资源、环境和社会经济等要素高度密集的综合体，因此城市的发展变化一直是众多学者们关注的焦点。随着遥感技术的发展，遥感影像分辨率不断提高，如何高效准确地从影像上提取城市扩展信息成为城市扩展研究的一个重要领域。传统的城市信息提取方法是利用基于像元的分类算法直接从影像上分类，这种分类方法在分类时仅用到了影像的光谱信息，分类结果受色彩的影响十分敏感，导致分类结果较破碎，出现“椒盐现象”。面向对象的遥感信息提取技术的出现，有效克服了传统基于像元分类的缺陷，其最小分类单元是同时具有光谱、形状、纹理及拓扑等多种特征的影像对象，分类时可以利用多种对象特征对影像分类。 本文以成都市中心城区为研究对象，分别从1992年、2001年、2003年和2005年4个时期的多源多光谱多分辨率遥感影像上提取出3个时段内的城市扩展信息，分析13年间成都市中心城区的城市扩展情况。着重探讨了基于时序遥感影像和面向对象分类技术的城市扩展信息提取方法，提出了两种不同的提取城市扩展信息的具体方法。一种方法是分别提取前后两个时相的城市用地信息，再利用前一时相的城市分类数据对后一时相的分类数据进行分割进而提取出城市扩展信息。另一种方法是利用特征融合影像、NDVI及不同尺度的纹理影像的组合，直接分割提取城市扩展信息。两种方法中均利用多尺度分割得到的影像对象建立了对象的多层次结构体系，有效提取出城市用地、农业用地、河流及坑塘等不同尺度的地物信息，取得了良好的分类效果。本文还进一步采用地类精度评价和面积精度评价两种方案对信息提取结果进行了定量的评价，同时与传统基于像元的分类结果比较，证明了本文提出的城市扩展信息提取方法在城市扩展信息提取中的有效性。 最后，利用1992-2001年、2001-2003年和2003-2005年3个时段内的城市扩展数据，分析成都市中心城区的城市扩展时空变化。选取城市用地年扩展速率、年扩展贡献率、年平均扩展强度指数等指标，对成都市中心城区的城市扩展情况进行分析。得出结论：1992-2005年13年间，成都市中心城区处于城市扩展快速增长时期，城市主要朝西部、西北、西南3个方向扩展。 |
| 英文摘要 | The city is a complicated aggregate which consists of many elements such as population, resources, environment and socio-economy etc. Its development has always been the focus of attention of scholars. With the development of remote sensing technology, the spatial resolution of remote sensing images continuously improves. How to efficiently and accurately extract the information of urban expansion from the images becomes a significant research field of urban expansion. The traditional method of urban expansion information extraction is based on the pixels of images for classification and thus there exist numerous peppers and salts in the classification result due to it merely taking use of the spectral feature of images in the process of classification. In order to improve the classification results, the technology of object-oriented information extraction comes into being, which is based on the image objects. The image objects generated from the pixels of images have many features such as shape, texture, topology and so on besides spectral feature and they can be used in the process of classification. In this paper, taking Chengdu central area as the experimental area, the urban expansion information was extracted from the multi-source, multi-resolution and multi-spectral remote sensing images in the periods of 1992-2001, 2001-2003 and 2003-2005, and then its expansion during 13 years was analyzed. The methods of urban expansion information extraction with the temporal and sequential remote sensing images and the technology of object-oriented classification were mainly discussed, and two different concrete methods were put forward. One is that first, the urban information were extracted separately from the images of the former year and the latter year; then the classification result of the latter year was segmented again by the classification result of the former year; last, the urban expansion information based on the segmentation was extracted. The other is that the combination of the images of feature fusion, NDVI and textures with different scales were used to extract the urban expansion information. In both of two methods, using multi-resolution segmentation to construct hierarchical network of image objects was taken, and urban, rural and water were extracted successfully. The accuracy of the classification result was evaluated in two facts of type and area. Compared with traditional classification based on pixels, the method in this paper is much better. Finally, with the classification results in the periods of 1992-2001, 2001-2003 and 2003-2005, the temporal and spatial expansion of Chengdu central area was analyzed by using indicators of the growth rate, the contribution rate and the average expansion intensity index etc. The result shows: from 1992 to 2005, Chengdu central area grows fast and mainly expands towards the west, northwest and southwest. |
| 学科主题 | 摄影测量与遥感技术 |
| 语种 | 中文 |
| 公开日期 | 2010-10-14 |
| 分类号 | TP3;TP1 |
| 源URL | [http://ir.imde.ac.cn/handle/131551/2186]  |
| 专题 | 成都山地灾害与环境研究所_成都山地所知识仓储（2009年以前） |
| 推荐引用方式 GB/T 7714 | 兴光. 基于时序遥感影像和面向对象分类技术的城市扩展信息提取研究[D]. 成都. 中国科学院水利部成都山地灾害与环境研究所. 2008. |
入库方式： OAI收割
来源：成都山地灾害与环境研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。