面向对象的居民地变化检测方法研究

发布时间：2018-01-15 11:14

本文关键词：面向对象的居民地变化检测方法研究　出处：《中南大学》2013年硕士论文　论文类型：学位论文

【摘要】：居民地数据是基础地理信息的核心要素之一。在我国城市化的发展进程中,居民地数据变化频繁。利用遥感技术及时、准确地发现、确定居民地变化对灾害评估、城市扩张、环境变化、空间数据更新等有着重要意义。因此,本文研究和发展了一种从多时相遥感影像中提取居民地变化信息的方法,主要研究工作如下： (1)分析了遥感影像多尺度分割的原理,选取不同分辨率的影像进行分割,并对分割过程中的尺度选择进行了研究和比较实验。 (2)提出了一种基于结构相似度的居民地自动提取算法。该算法首先将选取的居民地样本特征信息存入到MySQL数据库中,并借助结构相似度指数,避免了相似样本的重复存储,使得数据库中的样本相关性很小；然后将分割后的影像对象的灰度和纹理特征依据结构相似度与样本进行比较,即可得出居民地类别。另外,由于样本存在于数据库中,可以不断重复使用,提高了样本选择效率。 (3)提出了一种基于面积重叠比率的变化图斑判断方法。该方法将两时相影像提取的居民地图斑进行叠加操作,通过面积重叠比率的大小来发现变化图斑并确定变化类型(未变化、局部变化、新增、消失)。 (4)以Visual Studio2008为开发工具,结合Arcgis Engine9.3以及ENVI/IDL,开发了居民地变化检测的原型系统,并对本文所提出的方法进行实验验证,同时还与另外两种检测方法进行对比。实验结果证明本方法可以较好地提取居民地变化图斑,检测精度较高,而且要优于另两种检测方法。
[Abstract]:Residential data is one of the core elements of basic geographic information. In the process of urbanization in China, the data of residential land change frequently, and the remote sensing technology is used to find it in time and accurately. The determination of residential land change is of great significance to disaster assessment, urban expansion, environmental change, spatial data updating, and so on. This paper studies and develops a method to extract the information of land change from multitemporal remote sensing image. The main research work is as follows: 1) the principle of multi-scale segmentation of remote sensing image is analyzed, and the different resolution images are selected for segmentation, and the scale selection in the segmentation process is studied and compared. In this paper, we propose an automatic extraction algorithm based on structural similarity. Firstly, the selected resident sample feature information is stored in the MySQL database, and the structural similarity index is used. It avoids the duplicate storage of similar samples, which makes the correlation of samples in database very small. Then the grayscale and texture features of the segmented image objects are compared with the samples according to the structural similarity, and the categories of residents can be obtained. In addition, because the samples exist in the database, they can be used repeatedly. The efficiency of sample selection is improved. In this paper, a new method based on area overlap ratio is proposed, which overlay the population map spot extracted from 2:00 image. The size of the area overlap ratio is used to find the change pattern and determine the change type (no change, local change, new, disappear). 4) using Visual Studio2008 as the development tool, combining Arcgis Engine9.3 and ENVI/IDL. The prototype system of residential change detection is developed, and the method proposed in this paper is verified by experiments. The experimental results show that this method can extract the change pattern of residents better, the detection accuracy is higher, and it is better than the other two detection methods.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP751;P208

【参考文献】