【作者】 费鲜芸；

【导师】 张志国；

【作者基本信息】 山东农业大学 ， 土壤学， 2006， 博士

【摘要】 绿地不仅是城市的净化器,具有生态功能,同时也有着重要的社会功能。随着经济增长和社会发展,绿地的规划、建设、评价、监测和管理越来越受到重视。利用遥感技术获取绿地信息成为快速、客观、准确的城市生态监测、评价、规划和管理的重要手段。目前可以利用的高分辨率遥感数据资料越来越多,高于1m分辨率航天遥感影像和航空遥感影像已开始应用到资源调查和测图中。本文以城市不同绿地类型信息提取为目标,从几何校正、数据融合、分类方法等方面着手进行研究,探索研究不同高分辨率遥感数据、不同信息提取方法在城市绿地信息提取中的应用,并在此基础上进行绿地功能综合评价研究。为城市绿地信息提取适宜遥感影像尺度的选择、图像预处理、图像分类及评价提供技术和方法支持。论文首先研究了利用2D二次多项式校正城区不同高分辨率遥感影像所能达到的面积和点位精度。分别用不同数量的(25和6)GPS地面测量点和从1∶10000地形图中获取的控制点校正SPOT5影像;利用6个GPS控制点校正IKONOS影像;利用6个GPS控制点校正QUICKBIRD影像,并利用另外的GPS地面点作为检核点,对校正后的影像能达到的点位和面积精度分别进行研究。结果表明:(1)在地形起伏不大的城区,采用二维二次多项式的方法校正SPOT影像(2.5m分辨率),用较少数量的GPS控制点(6个)即可得到较好的点位精度;(2)如果用地形图校正影像,需要足够数量且分布均匀的控制点来获取符合要求的点位精度。(3)利用GPS控制点和地形图控制点校正SPOT5影像均可获得较好的面积精度,满足绿地调查等专题制图需要。(4)校正后影像上点位和面积精度存在复杂的关系,不能简单地利用误差理论由点位精度推导面积精度。(5)在地形起伏不大的城区,用6个GPS控制点,采用二维二次多项式模型,校正QUICKBIRD和IKONOS影像均可获取较好的点位和面积精度,与SPOT5数据相比二者点位精度明显较高,但并不表现明显高的面积精度。第二部分是在影像校正的基础上,针对城区植被特点,研究高分辨率遥感数据的融合方法。同一融合方法不同应用目的融合效果不同,本文采用主成分分析、HIS变换以及基于小波变换的主成分分析和HIS变换四种方法对SPOT5全色影像和多光谱影像进行融合,并结合城市植被制图特点对融合结果进行质量评价。结果表明,基于小波变换的PCA和HIS变换融合法,光谱保持能力最好,但是空间结构较差,有明显的分块效应,不适于城市植被零星分布的特点。主成分分析既有较好的空间结构特征,细小地物纹理清晰,同时又具有较好的光谱保持能力,最适合于城市植被制图研究。缨帽变换是针对植被提取进行的一种图像增强方法,变换后的各分量与植被、土壤等自然景观属性相联系,能更好地区分植被类型。针对城市植被特征,本文研究了基于缨帽变换的IKONOS全色和多光谱影像的融合方法,同时利用主成分分析方法对影像进行融合,分别在整景影像和植被区对融合结果进行对比分析。结果显示在该实验区,对整景影像和植被区域,从目视效果和定量指标两个方面进行评价,缨帽变换融合结果都优于主成分析结果。第三部分是基于高分辨率遥感影像的绿地信息提取方法的研究。建城区绿地面积是指公共绿地、居住区和单位附属绿地、防护绿地、生产绿地、风景林地、道路绿化用地等六类绿地面积之和。建城区绿化覆盖面积是指城市绿化种植中的乔木、灌木、草坪等所有植被的垂直投影面积及水面之和,其中建成区范围内的耕地不统计在内。绿地面积具有社会属性,难以利用常规的以光谱特征为基础的分类方法进行提取,本文利用面向对象的影像分析技术,以20cm航空影像为数据源,利用光谱特征、长宽比特征、距离特征及语义关系等,对城市道路绿地和零星绿地进行了自动提取,结果表明提取精度精度分别达到了76. 56%和88.45%以上。为研究绿化覆盖面积的提取,本文以SPOT5影像为数据源,在分析不同植被类型、水面、裸土、耕地、道路和居民区的光谱特征的基础上,利用多光谱影像计算两个指数模型(G-R)/(G+R)和(G+SWIR)/80,并结合多光谱的第四波段,采用基于耕地分区的决策树法,对城市绿化覆盖面积进行提取,结果表明与利用监督分类对融合后的影像进行分类相比,提取精度明显提高。第四部分是遥感尺度对绿地信息提取精度的影响。遥感影像信息提取的不确定性和尺度是遥感信息科学中两个密切相关的概念,研究影像信息不确定性就需要了解不同的尺度对信息提取的精度有何影响。由于城市绿地林相复杂,且有大量的零星绿地存在,而且许多绿地分布在高层建筑物之间,如果分辨率低,则零星绿地不能充分表现,如果分辨率过高又会出现阴影和遮挡严重的现象,本文对三种常用高分辨率遥感影像SPOT5、IKONOS和QUICKBIRD,通过目视解译方法进行绿地信息提取,分析其绿地信息提取的精度。结果表明:较高分辨率影像获取面状和带状绿地信息,精度优于较低分辨率影像;较高分辨率遥感影像提取零星小绿地的精度明显较高;在以7层以下建筑物为主的城市区域,绿地与道路的结构特点决定,阴影和遮挡造成的绿地信息丢失,较高分辨率遥感影像与较低分辨率遥感影像差别不大。所以适应城市绿地信息提取特征,三种影像中,QUICKBIRD影像最适合绿地调查。第五部分是以RS和GIS为基础,对城市主要绿地类型,即公共绿地和道路绿地,进行绿地功能综合评价。为了对城市公共绿地进行评价,本文根据公共绿地的功能特点,按照评价体系可操作性和可比性要求,选择单位公园绿化覆盖率、面积大小、公园服务半径内的居民区和单位面积比例、服务半径内道路面积比例作为评价因子。论文以山东省泰安市为研究区,利用基于植被指数的决策树分类法结合ArcGIS空间分析功能,从SPOT5影像中获取评价因子数据,利用多级模糊评判方法对研究区城市公共绿地进行评判。评价结果为泰安市公共绿地的生态功能和社会功能综合得分为77.449分,等级标准为良好。为对研究区道路绿地的功能进行评价,需用三步生成所需要的道路系统图和道路绿化覆盖图。首先在SPOT5(2.5m)假彩色遥感影像中手动绘制道路中线,并根据每条道路的宽度分别做缓冲区分析,生成城市道路系统图;再通过对SPOT5多光谱影像计算植被指数,并利用决策树分类法获取城市绿化覆盖图;最后通过对整个城市的绿化覆盖图与道路系统图进行空间叠加分析,生成城市道路绿化覆盖图。在此基础上计算道路的绿化覆盖率及道路绿地的连接度指数,结果表明研究区总体有较好的道路绿化覆盖率,但大部分道路的连接度较差。更多还原

【Abstract】 The green space system plays the important ecological and social roles in unban area. With the development of economy and society, more and more attantion are payed to urban green space planning, construction, evaluation, monitoring and management. It becomes a basic, efficient and precies method to obtain the urban green space information by using remote sensing (RS) technology. There are several high spatial resolution RS images available now for urban green space surveying, including aerial and satellite images with less than 1m of the spacial resolution. In this paper, image correction, data fusion, classification related with urban green space system were studied. Based on those studies, functions of main types of green spaces were evaluated.The first research objective is to probe the area and position precisions of different urban high resolution images corrected by using 2D 2sd polynomial in urban region. Different precision (acquired from GPS field surveying and 1:10000 topography maps) and number (25 and 6) of ground control points (GCPs) were used to correct the SPOT5 image; six GPS GCPS were used to correct the IKONOS image; six GPS GCPS were used to correct the QUICKBIRD image. Some other GPS points were used as check points (CPs) to check the position and area precisions of the corrected images. The results showed that: (1) For urban SPOT5 image, through 2D 2sd polynomial function by using 6 high precision GPS GCPs, better position precision was obtained in flat urban area. (2) Satisfied position precision could be reached by using 1∶10000 topography maps if GCPs are enough and their locations were well distributed. (3) Satisfied area precision was reached on the images corrected by GPS GCPs and map GCPs for urban green space system surveying. (4) The relationship between position and area errors is complex on the corrected images, so the area errors cannot be deduced by point errors according to theory of errors.(5)In the flat urban region, better position and precision were obtained through 2D 2sd polynomial function using 6 high precision GPS GCPs for IKONOS and QUICKBIRD images.The second research objective is to probe the data fusion for urban vegetation mapping based on image correction. The effect is not consistent when different fusion methods were used for different purposes. In this study, SPOT5 images were fused using four fusion methods, including PCA, HIS, PCA based on wavelet transformation, and HIS based on wavelet transformation. The conclusions were that the image fused by PCA based on wavelet transformation reached better spectral quality, but the spatial construction is not well. Because the location characteristics of urban vegetation is fragmental, so this method is not suitable for vegetation information obtaining. Better image spatial construction and spectral quality were obtained by PCA fusion. Because texture of trivial objectives is distinct, this method is best for vegetation mapping among these four methods.Tasseled Cap (KT) transformation is useful to distinguish different types of vegetation. The bands after transformation are correlated with nature landscape such as vegetation, soil. The IKONOS data was fused with KT and PCA methods. The fusion results were analyzed and compared within full image and vegetation subset. The result showed that the KT fusion method is better than PCA in full image and vegetation subset in the study area. The fusion results are differential with different KT transformation matrix, and the method of fusion should be chosen according to study area, data, and classification methods.The third research objective is to probe the methods of urban green space classification. The area of urban green space is the sum of six types of green space, including public green space, residential green space, nursries, recreation forest, green area for environmental protection, roadside green spaces. Green coverage is the area of all growing planting projection, including tree, shrub and grass. The cultivated lands in the urban region are not belongs to the green space.The green space area has social attribute, so it is difficult to get the each type classification information by spectral characteristics. In this paper, based on the image analysis technology of oriented-objective, spectral characteristics combined with ration of length and width, distance characteristics was used to get the information of roadside green space by computer automatic classification on 20cm resolution aerial image. The results showed that classification precision is better than 76. 56%.In order to obtain the green coverage information, the spectral characteristics of main classification objectives including various type vegetation, water body, bare soil, cultivated land, road area, and building of SPOT5 multi-spectral image were analyzed. Based on spectral analysis, two index models, (G ? R)/(G+R) and (G + SWIR)/80 were calculated. Decision tree algorithm classification method combining cultivated land distributions was used to get the green coverage by using two index models and the fourth band of multi-spectral image. Contrast to the supervised classification on fused image, the classification precision is obviously higher.The fourth research objective is to probe the relationship between the green space classification and scale. The uncertainty between RS image information acquiring and the size of scale is two highly related conceptions. It is necessary to study the scale impact to information acquire precision caused by image information uncertainty. In urban area, the structure of vegetation is complicated and the location of green space is fragmental. In the same time much green space is located in the shade of high buildings. Fragmental green space can not be fully reflected on lower resolution image, but there are more shadows and the shade of high buildings with higher resolution images. In this paper the green space information of same region was obtained by visual interpretation from SPOT5, IKONOS, and QUICKBIRD images and the classification precision were compared with. The results showed that: (1) For big size green space, higher resolution images have the better classification precision than lower. (2) Higher resolution images have obviously fine precision for fragmental green space classification. (3) In the urban region that most buildings less than 7 levels, the precision differences caused by the green space information lose for the shadow and shade of building is not obvious. So QUICKBIRD image is most suitable for urban green space surveying among three high resolution images.The fifth research objective is to probe the synthetical methods of the function evaluation of main green space types. In order to evaluate the function of public green space, four factors were chosen according to public green space function characteristics and evaluation requirement, which are percentage of greenery coverage, size of area, ratio of residential area within service radius, ratio of road area inside service radius of each public green space. Tai’an, Shandong, was chosen as the study case. Evalation factors data were obtained by RS decision tree algorithm classification based on vegetation indexs and ArcGIS spatial analysis methods from SPOT5 image. Fuzzy multi-level synthetic evaluation was used to evaluation the public green space of entire urban region. The result of evaluation is that the synthetical grade of public green space ecological and social function of Tai’an is 77.449, which indicates fine quality according to evaluation standard.In order to evaluate the ecological function of roadside green space in study region, there are three steps to get urban road system map and road green coverage map. (1) Urban road system map was obtained through buffer analysis by using middle lines of urban roads drawn by manual methods on SPOT5 2.5m resolution false color image. (2) Urban green coverage map was obtained by decision tree algorithm classification based on vegetation index on SPOT5 multi-spectral image. (3) Road green space coverage map was obtained by spatial overlap analysis by using the road system map and urban green space coverage map. Based on these two maps, percentage of road greenery coverage and road green space connectivity was calculated to evaluate the quality of the road green space system.更多还原