基于多源遥感数据的建筑工程资产投资态势监测

发布时间：2018-03-03 04:29

本文选题：固定资产投资　切入点：遥感监测　出处：《中国地质大学(北京)》2016年博士论文　论文类型：学位论文

【摘要】：固定资产投资对城市景观格局变化、城市经济发展有举足轻重的作用,其动态监测技术研究也是近年来的研究热点。本文选取了河南省中牟县和重庆市北碚区两个研究区,以固定资产投资与城市扩张、投资项目特征提取方法、城市景观格局的驱动机制作为切入点监测不同尺度的投资态势。主要研究内容与结论如下:(1)通过构建1990年至2010年的全国各区域全社会固定资产投资面板数据并进行系数修正,结合遥感人工解译的1990年至2010年的632个城市的主城区扩张面积空间分布,建立二者的回归模型,揭示了宏观尺度固定资产投资规模的趋势及分布规律,提出利用遥感监测建成区扩张监测投资规模的方法。揭示了固定资产投资存在的“虹吸”效应。数据说明,随着中国改革的不断推进,投资规模与扩张面积的相关性持续减弱,投资对经济的驱动力权重有所下降。单位扩张面积的固定资产投资额这一土地投入指标显示,中国城市的扩张集约节约效果显著。(2)提出了针对高分辨率、超高分辨率遥感数据的投资特征提取方法。实验证明,HSD特征训练得到的随机森林机器学习以提取目标建筑物的方法为从超高分辨率遥感影像中剔除土壤这一方向提供了一种鲁棒性较强的方法。而且,随机森林分类器移植性远远高于传统方法,可用于提取投资监测领域的多种监测目标。此外,针对塔吊这一特殊的投资项目在建配套设备,使用数学形态学和几何特征相结合的算法进行精确的定位和数量提取,并进行了实验验证。(3)建立了适合遥感投资监测的包括15个类别的投资项目监测体系。针对可监测项目构建了特征指数BBI及IPBI,应用面向对象的分类算法,结合光学纹理等多种特征值,对彩板房等临时建筑物等对象进行解译,有效获取了微观尺度的投资热点分布。(4)通过对平原研究区中牟县和山地研究区北碚区的近15年四个时相的土地利用景观类别动态的监测,构建基于平原与山区的景观格局指数框架,对不同固定资产投资结构对景观格局变化的驱动差异性进行探索性分析。
[Abstract]:Fixed asset investment plays an important role in the change of urban landscape pattern and the development of urban economy. The research on dynamic monitoring technology is also a hot topic in recent years. This paper selects Zhongmou County in Henan Province and Beibei District in Chongqing as two research areas. The method of extracting the characteristics of fixed assets investment and urban expansion, The driving mechanism of urban landscape pattern is used as the starting point to monitor the investment situation of different scales. The main research contents and conclusions are as follows: (1) by constructing the data of fixed assets investment panel of the whole society from 1990 to 2010 in the whole country and modifying the coefficient, Combined with the spatial distribution of the expansion area in 632 cities from 1990 to 2010, the regression model of them is established, and the trend and distribution law of the scale of fixed asset investment in macro scale are revealed. This paper puts forward a method of using remote sensing to monitor the investment scale of the established area, and reveals the siphon effect of the fixed asset investment. The data show that the correlation between the investment scale and the expansion area continues to weaken with the development of China's reform. Investment has a lower driving force on the economy. Investment in fixed assets per unit expansion area, a land input index, shows that the intensive expansion of Chinese cities has significant savings. An investment feature extraction method for ultra-high resolution remote sensing data. It is proved by experiments that the method of random forest machine learning based on HSD feature training to extract target buildings is to remove soil from ultra-high resolution remote sensing images. Provides a robust approach. And, The transplantability of stochastic forest classifier is much higher than that of traditional methods, and it can be used to extract various monitoring targets in the field of investment monitoring. Using mathematical morphology and geometric features of the algorithm for accurate location and quantity extraction, The monitoring system of 15 kinds of investment items suitable for remote sensing investment monitoring is established. The feature index BBI and IPBI are constructed for the monitored projects, and the object-oriented classification algorithm is applied. Combining with the optical texture and other characteristic values, the objects such as temporary buildings, such as color plate houses, are interpreted. The distribution of investment hot spots on the micro scale is obtained effectively. The dynamic monitoring of land use landscape types in the past 15 years in Zhongmou County, plain research area, and Beibei, mountainous research area, is carried out. The landscape pattern index framework based on plain and mountain area is constructed to analyze the driving difference of different fixed asset investment structure on landscape pattern change.
【学位授予单位】：中国地质大学(北京)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F283

【参考文献】