面向地表覆盖数据更新的变化检测方法研究

发布时间：2018-06-15 02:05

本文选题：地表覆盖更新 + 变化检测　；参考：《武汉大学》2014年博士论文

【摘要】：地表覆盖数据是全球变化研究、地理国情普查和生态资源管理等工作的重要基础信息。自然和人为因素的影响使地表覆盖不断发生着变化,因此地表覆盖数据的现势性是衡量其使用价值的重要标志之一。变化检测是地表覆盖更新的关键技术,目前已有多种变化检测算法,如变化向量分析法、差值法、相关系数法和夹角余弦法等,但是这些算法大多从量值的角度进行变化检测,容易受多种因素的干扰产生伪变化。本文从光谱(或NDVI)曲线形状的角度出发,利用斜率描述光谱(或NDVI)的曲线形状,通过斜率计算将变化检测转化到斜率空间,基于斜率差异进行变化检测,本文的主要研究包括以下四个方法方面： (1)描述了地表覆盖及变化的概念和重要性,在对全球尺度和区域尺度地表覆盖制图产品分析的基础上,指出变化检测是目前地表覆盖更新的主要手段。评述国内外变化检测的研究现状,分析目前变化检测存在的主要问题,即容易受非地表覆盖变化因素的影响产生伪变化。根据伪变化产生的原因,将其分为随机干扰因素造成的伪变化和季相差异因素造成的伪变化两类。针对变化检测存在的问题,提出本文的研究思路,并对本文章节安排进行介绍。 (2)针对随机干扰因素产生的伪变化问题,提出了基于光谱斜率差异的变化检测方法(Spectral gradient difference, SGD)。为了定量化描述Landsat光谱曲线的形状,对光谱斜率进行定义和计算,比较不同时期光谱曲线的斜率差异得到变化强度影像,利用双窗口变步长方法设置阈值,确定变化区域。针对变化区域,从斜率走势和量值两方面构建SGD链模型,描述不同变化类型的光谱斜率变化规律,组建变化规律知识库,通过变化像素的链和知识库的匹配确定出变化类型。最后,基于模拟实验和Landsat影像实验对变化检测算法进行验证,并和传统的变化检测方法进行精度对比。 (3)针对季相周期因素造成的伪变化问题,提出了基于NDVI斜率差异的变化检测方法(NDVI gradient difference, NDVI-GD)。通过Landsat和MODIS数据融合生成30米的NDVI斜率数据,利用NDVI斜率描述地物NDVI曲线的宏观走势,揭示地物的生长周期,通过比较NDVI斜率差异检测变化区域。在变化区域确定方面,基于斜率特征和季相特征构建NDVI斜率链,用以刻画地物间变化的季相周期规律,通过变化像素和链模型知识库的匹配确定变化类型。最后,利用陕西省礼泉县两期不同季相的影像进行NDVI-GD变化检测实验,通过实验证明了NDVI-GD能有效解决季相差异造成的伪变化问题。 (4)在上述研究内容的基础上,针对大范围地表覆盖类型多样且伪变化情况复杂的问题,提出了基于SGD和NDVI-GD的综合变化检测方法,以陕西省为实验区域,根据每景影像获取时间在季节上的差异,选取相应的变化检测方法,最后将陕西省2000基准年的地表覆盖数据更新到2010基准年。
[Abstract]:Surface cover data is an important basic information for the research of global change, the survey of geographical conditions and the management of ecological resources. The influence of natural and human factors makes the surface cover change constantly. Therefore, the present situation of surface cover data is one of the important indicators to measure its use value. Change detection is the key technology of surface cover renewal. At present, there are many change detection algorithms, such as change vector analysis method, difference method, correlation coefficient method and angle cosine method, etc. It is easy to be disturbed by many factors to produce false changes. In this paper, from the point of view of the shape of the spectral (or NDVI) curve, the slope is used to describe the curve shape of the spectrum (or NDVI), the change detection is transformed into slope space by slope calculation, and the change detection is carried out based on the slope difference. The main research contents in this paper are as follows: (1) the concept and importance of surface cover and its change are described. Based on the analysis of global and regional scale surface cover mapping products, It is pointed out that change detection is the main means of surface cover renewal at present. This paper reviews the research status of change detection at home and abroad, and analyzes the main problems existing in change detection at present, that is, it is easy to be affected by non-surface cover change factors to produce pseudo-change. According to the reasons of pseudo-variation, it can be divided into two categories: the pseudo-variation caused by random disturbance and the pseudo-variation caused by seasonal difference. In view of the problems existing in change detection, this paper puts forward the research ideas, and introduces the chapter arrangement of this paper. A method for detecting spectral slope difference based on spectral slope difference is proposed. In order to quantitatively describe the shape of the Landsat spectral curve, the spectral slope is defined and calculated, and the intensity image is obtained by comparing the slope differences of the spectral curve in different periods. The threshold value is set by using the method of variable step size of two windows to determine the region of change. In this paper, the SGD chain model is constructed from slope trend and quantity to describe the variation law of spectral slope of different change types, to establish the knowledge base of variation law, and to determine the change type by matching the chain and knowledge base of changing pixels. Finally, based on the simulation experiment and Landsat image experiment, the change detection algorithm is verified, and compared with the traditional change detection method. A method of NDVI gradient difference detection based on slope difference is proposed. Landsat and MODIS data are combined to generate 30 meters of NDVI slope data. NDVI slope is used to describe the macroscopic trend of NDVI curve of ground objects, and the growth cycle of ground objects is revealed. The variation regions of NDVI slope difference detection are compared. The slope chain of NDVI is constructed based on slope feature and seasonal feature to describe the seasonal periodic law of the change of ground objects. The change type is determined by matching the changing pixels and the chain model knowledge base. Finally, the NDVI-GD change detection experiment is carried out by using the images of different seasons in Liquan County, Shaanxi Province. It is proved that NDVI-GD can effectively solve the problem of pseudo-variation caused by seasonal phase difference. In order to solve the problem of diverse cover types and complex pseudo-variation in large scale, a comprehensive change detection method based on SGD and NDVI-GD is put forward. Taking Shaanxi Province as the experimental area, according to the seasonal difference of the acquisition time of each scene image, The corresponding change detection method is selected and the surface cover data of 2000 base year in Shaanxi Province are updated to 2010 base year.
【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：P237

【相似文献】