基于多源夜间灯光遥感影像的多尺度城市空间形态结构分析

发布时间：2017-12-31 14:00

本文关键词：基于多源夜间灯光遥感影像的多尺度城市空间形态结构分析　出处：《华东师范大学》2017年博士论文　论文类型：学位论文

【摘要】：城市作为人类活动的主要集聚地,城市的出现和发展受到了人类活动、自然环境以及社会经济活动的影响,同时其发展和扩张也同样影响着自然环境以及人类活动的模式。因此,对于城市的研究,尤其对城市空间结构的研究,是一个与人类切身利益相关且极具意义的研究方向。随着地理信息系统与遥感技术的日益发展,尤其是夜间灯光遥感影像的出现,相关技术被多次用于城市空间形态的研究,但仍存在一定不足,值得做进一步研究。首先,获取长时间序列的城市建成区范围是研究城市空间结构的基础。但目前已有的城市建成区数据在时间上不连续,且不同产品数据所认定的"城市区域"存在差异,导致不同数据间的不可比。其次,在城市空间结构识别的研究中,多是使用统计调查数据进行城市中心的识别工作,鲜有基于遥感技术识别城市中心以及城市中心间相互关系的研究。最后,传统城市居住空间结构大多以住房价格作为主要指标,但住房价格并不能真实反映该住房的使用情况。因此有研究对此提出了质疑,认为住房价格并不一定适于研究城市居住空间结构。城市住房空置率已被证明与住房价格呈负相关的关系,且能够反映住房的真实入住情况,故认为其可被用于城市居住空间结构的研究。但受限于当前仅有少数国家或地区统计或公布了住房空置率信息,且其统计尺度相对较大,不利于研究其空间分布模式。通过分析和归纳国内外已有的研究情况,提出了基于夜间灯光遥感影像的多空间尺度城市空间形态结构识别与分析方法,内容涵盖了长时间序列的全球城市建成区提取、城市空间结构识别以及城市居住空间结构探究。本文的主要研究内容和成果如下:(1)本文将"城市区域"定义为基于土地利用的城市建成区,并提出了基于DMSP-OLS(Defense Meteorological Satellite Program-Operational Linescan System)稳定夜间灯光数据以及 MODIS(Moderate-resolution Imaging Spectroradiometer)标准产品数据的区域生长支持向量机-双向马尔科夫随机场模型,提取了 2000年至2012年全球城市建成区范围。首先,详细介绍了利用区域生长支持向量机提取初始全球城市建成区范围的过程,随后利用双向马尔科夫随机场模型,融合了遥感影像的空间邻域及时间邻域信息,对初始城市建成区提取结果进行逐步优化和校正。通过与已有全球土地覆盖产品数据,证实了当前人类活动范围不仅仅局限于不透水层,已渗透至部分非不透水层区域(例如城市森林等)。通过与2000年、2005年以及2010年Landsat7ETM+多波段遥感影像进行精度验证,本研究提取的城市建成区范围总体精度和Kappa系数均处于较高水平,三个时相的总体精度(Kappa 系数)分别为 86.75%(Kappa:0.72)、87.93%(Kappa:0.75)以及88.39%(Kappa:0.77)。结果表明,基于区域生长支持向量机-双向马尔科夫随机场(RSVM-BMRF)模型能够有效地实现大空间尺度下的长时间城市建成区提取,且在时间序列上具有更稳定的表现。(2)本研究将夜间灯光数据视为反映人类活动聚集程度的连续场,并将地理要素与城市要素进行类比,包括将城市中心识别转化为在夜间灯光数据中寻找"山峰"的过程以及将地形学中的"坡度"应用至夜间灯光数据以表征人类活动的变化情况,最终利用局部等值线树算法提取多株等值线树来反映研究区内夜间灯光亮度的变化情况,实现对城市空间结构的探究。本研究以中国上海为研究区,利用NPP-VIIRS(Suomi National Polar-Orbiting Partnership-Visible Infrared Imaging Radiometer Suite)夜间灯光数据成功识别出了 33个城市中心的位置、空间范围以及层级关系。通过量化城市中心的形态指标与夜间灯光指标,有效地分析了上海市各个城市中心的发展情况及多中心程度。基于等值线树的层级结构,使得本研究能够获取到城市中心的层级关系,以满足不同空间尺度下研究城市空间结构的需求。通过将地形学理论与城市地理学相结合,为今后研究城市空间结构提供了一种新的思路。(3)本研究提出了一种基于NPP-VIIRS夜间灯光数据以及美国土地覆盖数据估算城市住房空置率的模型以探究城市居住空间结构的思路。在剔除了非住宅要素的灯光亮度影响以及削减了夜间灯光像元内不透水层比例不一致所带来的灯光亮度差异影响的基础上,结合统计模型,估算了研究区内住房满置情况下的夜间灯光亮度值,最终实现住房空置率的估算以及城市居住空间结构的分析。受限于统计数据的获取,本研究仅以美国都会区为例。通过将住房空置率估算结果与统计调查数据进行对比,证明了该模型是可靠有效的(皮尔森相关系数为0.856,决定系数为0.734以及均方根误差为17.72%),可被用于填补其他国家或地区,或是在精细尺度上统计数据的空白。通过分析城市住房空置率分布模式,得出结论,美国总体居住区规划相对合理,尤其是在城市发展水平相对较高的区域。此外,自然要素(如山脉、海湾等)亦会对城市居住空间结构产生影响。
[Abstract]:City as the main gathering place of human activities, the emergence and development of city by human activities, the influence of natural environment and social economic activities, while its development and expansion also affect the natural environment and human activities. Therefore, the research for the city, especially the city spatial structure research is a research direction related to the vital interests of human beings and has great significance. With the development of geographic information system and remote sensing technology, especially the night lights of the remote sensing image, related technology has been used to study many city space form, but there are still some problems that need to be further studied. First of all, to obtain long time series of the built-up area of the city the range is the basis of the research on city spatial structure. But the existing city built-up area data in time is not continuous, city area and different product data identified by " "There are differences, resulting in different data is not comparable. Secondly, in the study of spatial structure of city identification, is the city center using the survey data of the identification work, little research on relationship between city center and remote sensing recognition based on the city center. Finally, the traditional city residential space structure in most of the housing prices as the main index, but housing prices can not reflect the use of the housing. Therefore studies have questioned this, that the housing price is not necessarily suitable for the research of city residential structure. City housing vacancy rate has been shown to have a negative correlation with the housing price, and can reflect the real situation in the housing, so that it can be used to study the city residential space structure. But limited to the current statistics, only a few countries or regions or announced the housing vacancy rate and the information. The statistical scale is relatively large, is not conducive to the study of the spatial distribution pattern. Through the analysis and summary of the existing researches at home and abroad, put forward the multi structure scale city space form space identification and analysis method of night lights based on remote sensing images, covering the content of long time series of the global city built-up area extraction, identification of city and city residential space structure the spatial structure of research. The main research contents and results are as follows: (1) the "city region" is defined as land use based on the city built-up area, and put forward based on DMSP-OLS (Defense Meteorological Satellite Program-Operational Linescan System) and MODIS stable night light data (Moderate-resolution Imaging Spectroradiometer) standard product data of the region growing support vector machine - bidirectional Markov random field model, from 2000 to 2012 The city built-up area. Firstly, introduces the support vector machine to extract the initial growth of global city built area process by area, followed by two-dimensional Markov random field model, spatial and time neighborhood information fusion of remote sensing images, the initial City built-up area extraction results were gradually optimized and by covering the product data and correction. The global land, confirms the current range of human activity is not limited to the impermeable layer, have infiltrated the non impervious surface area (e.g. city forest). By 2000, 2005 and 2010 Landsat7ETM+ to verify the accuracy of multi band remote sensing images, the extraction of the city built-up area are in the overall accuracy and Kappa coefficient a high level, the overall accuracy of phase three when (Kappa coefficient) were 86.75% (Kappa:0.72), 87.93% (Kappa: 0.75) and 88.39% (Kappa 0.77). The results show that based on region growing support vector machine - bidirectional Markov random field (RSVM-BMRF) model can effectively realize long time city space scale built area extraction, and has more stable performance in time series. (2) the night light data as a reflection of human activities together the degree of the continuous field, and the geographical elements and elements of the city including the city center analogy, recognition into night light data for the "peaks" and the "slope" topography will apply to night light data to characterize the changes of human activities, the final contour extraction algorithm based on local tree strains the contour tree to reflect changes in the study area the night light, on the realization of city space structure. In this study, China Shanghai as the study area, using NPP-VIIRS (Suomi National Polar -Orbiting Partnership-Visible Infrared Imaging Radiometer Suite) night light data successfully identified 33 out of the city center location, spatial scope and hierarchical relationship. Through morphological index quantification in the center of the city night lights and indicators, and effectively analyses the development situation of the city center of Shanghai city and multi center. The contour tree based on hierarchical structure that makes this study can get to the hierarchy of the center of the city to meet the city spatial structure research under the different spatial scales of demand. By combining the topographic theory and city geography, provides a new idea for the future research on city space structure. (3) this study proposes a NPP-VIIRS night light data and the land cover data to estimate the city housing vacancy rate model to explore the city residential space structure based on the idea of taking out. Based on the brightness of the light effect of non residential elements and reduces light brightness difference of night lights in a pixel impervious layer brought by the difference of the ratio, combined with the statistical model, the study area is full of housing under the night light brightness value estimation, finally realize the housing vacancy rate estimation and analysis of City residential space limited by the structure. To obtain statistical data, this research only in the United States metropolitan area as an example. The housing vacancy rate estimates were compared with the results of the survey data, it is proved that the model is reliable and effective (Pearson correlation coefficient is 0.856, the coefficient of determination was 0.734 and the root mean square error is 17.72%), can is used to fill in other countries or regions, or blank statistics in fine scale. Through the analysis of city housing vacancy rate distribution model, concluded that the overall planning of residential area is In particular, there are relatively high levels of urban development. In addition, natural elements (such as mountains, bays, etc.) will also have an impact on the urban residential space structure.

【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TU984.113;P237

【相似文献】