地表温度空间降尺度方法及其基于国产高分影像的应用研究

发布时间：2018-11-23 12:02

【摘要】：地表温度是表征气候变化的重要环境参数。卫星热红外遥感作为全球和区域地表温度获取的重要手段,其较低的空间分辨率(六十米到几十千米),无法满足高异质城市环境中复杂地表温度变化监测的需求。地表温度空间降尺度是融合中低分辨率热红外数据和高分辨率可见光/近红外数据,获取高分辨率地表温度的重要手段。国产高分一号、高分二号卫星高空间分辨率、高定位精度、高整星机动能力的优势,为地表温度空间降尺度的研究提供了良好的数据基础。然而,目前国内外地表温度空间降尺度研究中,估算高分辨率(米级)地表温度的研究较少,现有研究忽略了同类地物反射光谱和热特性存在变异性这一问题,特别是在高分辨率下,类内的变异性更大,极大影响了高分辨率地表温度估算的准确性。针对上述问题,本文基于国产高分一号(GF-1)、高分二号(GF-2)多光谱影像和Landsat-8热红外影像,围绕同类地物反射光谱和热特性存在变异性这一事实,引入面向对象的图像分析和多端元混合像元分解方法,开展地表温度空间降尺度的方法和应用研究。具体研究内容如下:(1)多端元对象混合像元分解方法。集成面向对象的图像分析和多端元混合像元分解方法的优势,针对高分辨率影像,提出基于影像分割和局部最小平均光谱角(Minimum Average Spectral Angle,MAS A)的多端元对象混合像元分解方法(Multiple Endmember Object Spectral Mixture Analysis,MEOSMA),通过解决传统多端元混合像元分解方法在端元光谱抽取和端元集优化方面的不足,进一步降低类内变异性对光谱混合分解的影响。(2)基于MEOSMA的地表温度空间降尺度方法。依托国产GF-1、GF-2多光谱影像和Landsat-8热红外遥感影像,基于MEOSMA和线性热混合模型,提出了MEOSMA-TM(Multiple Endmember Object Spectral Mixture Analysis and Thermal Mixing)地表温度空间降尺度方法,有效解决了同类地物反射光谱和热特性变异性大的问题,实现了高空间分辨率地表温度的估算,并分别与降尺度前地表温度和杭州市国家基准气象站实测地表温度数据进行比较,具有较好的准确性。(3)基于MEOSMA-TM的杭州市交通网络对城市地表温度场的影响研究。基于MEOSMA-TM方法计算杭州市主城区多个时相的高空间分辨率地表温度数据,分析杭州市主城区地表温度场的时空分布特征;利用网格分析、缓冲区分析、空间统计分析等方法,研究交通路网密度、不同等级道路、绿化形式、交通枢纽对城市地表温度的影响。研究成果提高了高分辨率遥感影像混合像元分解和高分辨率地表温度估算的精度,推动了国产高分辨率卫星影像在城市热环境研究中的应用,对于小尺度城市热环境的研究具有重要的意义,对于城市土地利用规划、城市生态环境管理、交通路网规划等相关领域具有重要的应用价值。
[Abstract]:Surface temperature is an important environmental parameter to characterize climate change. Satellite thermal infrared remote sensing as an important means of global and regional surface temperature acquisition, its low spatial resolution (60 meters to tens of kilometers), can not meet the needs of complex surface temperature monitoring in highly heterogeneous urban environment. The spatial downscaling of surface temperature is an important method to obtain high resolution surface temperature by integrating low and low resolution thermal infrared data with high resolution visible / near infrared data. The advantages of high spatial resolution, high positioning accuracy and high maneuverability of the satellites made in China provide a good data basis for the study of the spatial downscaling of surface temperature. However, there are few researches on estimating the surface temperature at high resolution (meter level) in the space downscaling study of surface temperature at home and abroad, and the problem of variability of reflectance spectrum and thermal characteristics of similar ground objects is neglected in existing researches. Especially at high resolution, the intra-class variability is greater, which greatly affects the accuracy of high-resolution surface temperature estimation. In view of the above problems, this paper bases on the multispectral images of domestic high score 1 (GF-1), high score 2 (GF-2) and Landsat-8 thermal infrared images, and revolves around the fact that the reflectance spectra and thermal characteristics of similar ground objects have variability. Object-oriented image analysis and multi-terminal element mixed pixel decomposition are introduced to study the downscaling of surface temperature space and its application. The main contents are as follows: (1) Multi-terminal object mixed pixel decomposition method. Integrating the advantages of object-oriented image analysis and multiterminal hybrid pixel decomposition, a new method based on image segmentation and local minimum average spectral angle (Minimum Average Spectral Angle,) is proposed for high-resolution images. MAS A) 's mixed pixel decomposition method (Multiple Endmember Object Spectral Mixture Analysis,MEOSMA) solves the shortcomings of traditional multiterminal object mixed pixel decomposition methods in endmember spectral extraction and endset optimization. The influence of intra-class variability on spectral mixing decomposition is further reduced. (2) the surface temperature spatial downscaling method based on MEOSMA. Based on the domestic GF-1,GF-2 multispectral images and Landsat-8 thermal infrared remote sensing images, based on the MEOSMA and linear heat mixing model, a spatial downscaling method of MEOSMA-TM (Multiple Endmember Object Spectral Mixture Analysis and Thermal Mixing) surface temperature is proposed. It effectively solves the problem of high variability of reflectance spectrum and thermal characteristics of similar ground objects, and realizes the estimation of surface temperature with high spatial resolution. And compared with the downscaling surface temperature data and the measured surface temperature data of Hangzhou National benchmark Meteorological Station, it has good accuracy. (3) the influence of Hangzhou traffic network based on MEOSMA-TM on the urban surface temperature field is studied. Based on the MEOSMA-TM method, the high spatial resolution surface temperature data of several time phases in the main urban area of Hangzhou are calculated, and the space-time distribution characteristics of the surface temperature field in the main urban area of Hangzhou are analyzed. By means of grid analysis, buffer analysis and spatial statistical analysis, the effects of traffic network density, different grade roads, greening forms and traffic hubs on urban surface temperature are studied. The research results improve the accuracy of mixed pixel decomposition and high-resolution surface temperature estimation of high-resolution remote sensing images, and promote the application of domestic high-resolution satellite images in the study of urban thermal environment. It is of great significance for the study of small scale urban thermal environment, and has important application value for urban land use planning, urban ecological environment management, traffic network planning and other related fields.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：P407

【相似文献】