高分辨率遥感影像的道路提取方法研究

发布时间：2018-02-08 21:29

本文关键词： 道路提取图像分割中值滤波数学形态学 Hough变换　出处：《西安科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着遥感影像的获取途径越来越多、分辨率越来越高,我们获取目标影像也越来越容易。但是目前对影像的信息提取还主要依赖于人工解译,这种方法比较劳神费力,而且解译的准确度与解译员的专业能力有很大的关系。因此,如何能够实现遥感信息快速准确的半自动化或全自动化的提取有很大的意义。本文主要研究高分辨率遥感影像中道路的半自动化提取,道路信息是重要的基础地理信息,也是城市中更新最快的基础性地理信息之一。道路的及时更新对城市规划与建设、交通管理、GIS数据获取、空间数据库更新、影像理解有着非常重要的作用。因此实现道路的快速准确提取很重要的意义。在对影像进行道路提取之前,我们首先要对影像进行一系列的预处理,主要包含图像分割与中值滤波处理。本次论文首先采用不同的分割方法分别对影像进行分割,选出最适合本次实验的分割方法,再采用中值滤波对分割后的图像进行滤波处理以去除噪声点。在预处理过程中,根据影像的特点与道路的提取情况,也会包含线性拉伸、连通域删除小面积对象等处理。论文主要研究数学形态学与Hough变换这两种方法在提取遥感影像中道路的准确性。数学形态学是一种用以分析几何形状与结构的数学方法,属于一种非线性算子,由一系列形态学代数运算子组成。数学形态学的基本运算有4个:膨胀(或扩张)、腐蚀(或侵蚀)、开启和闭合。数学形态学算法简单灵活、运算速度快、易于运行,并且能够在对图像优化,保留原有信息的基础上,凸显图像几何特征以便进一步分析。Hough变换是一种用于检测二进制图像中图形的方法,如用来检测直线、圆、椭圆等,最经典的应用就是直线检测。Hough变换对图像中的噪声、边缘特征不连续、地形地貌变形等都有很好的适应性,可靠性高,算法稳定有效。因此本次论文先分别采用数学形态学与Hough变换两种方法对高分辨率遥感影像中道路进行提取测试,分析算法的优缺点,评定其提取精度。最后结合数学形态学与Hough变换两种方法,以资源三号卫星原始影像图的全色波段影像为实验对象,提取其中的道路,并评定道路的提取精度。
[Abstract]:With more and more ways to obtain remote sensing image and higher resolution, we can obtain target image more and more easily. But at present, the information extraction of remote sensing image mainly depends on manual interpretation, this method is more laborious. Moreover, the accuracy of interpretation is closely related to the professional ability of interpreters. It is of great significance to realize the rapid and accurate semi-automatic or fully automated extraction of remote sensing information. This paper mainly studies the semi-automatic extraction of roads in high-resolution remote sensing images. Road information is an important basic geographic information. It is also one of the fastest updating basic geographic information in the city. The timely updating of roads is helpful to urban planning and construction, traffic management, GIS data acquisition, and spatial database updating. Image understanding plays a very important role. Therefore, it is very important to realize the fast and accurate road extraction. Before we extract the image, we must first do a series of preprocessing to the image. This paper first uses different segmentation methods to segment the image, and selects the most suitable segmentation method for this experiment. Then the median filter is used to filter the segmented image to remove the noise. In the preprocessing process, according to the characteristics of the image and the extraction of the road, it will also include linear stretching. The paper mainly studies the accuracy of the two methods of mathematical morphology and Hough transform in extracting roads from remote sensing images. Mathematical morphology is a mathematical method used to analyze geometric shapes and structures. A nonlinear operator consisting of a series of morphological algebraic operators. There are four basic operations of mathematical morphology: expansion (or expansion), corrosion (or erosion), opening and closing. Easy to run, and on the basis of optimizing the image and preserving the original information, the geometric features of the image can be highlighted for further analysis. Hough transform is a method for detecting graphics in binary images, such as detecting straight lines, circles, etc. Ellipse, the most classical application is line detection. Hough transform has good adaptability and high reliability to noise, edge feature discontinuity, landform deformation and so on. The algorithm is stable and effective. In this paper, two methods, mathematical morphology and Hough transform, are used to extract and test the road in high-resolution remote sensing images, and the advantages and disadvantages of the algorithm are analyzed. Finally, with the combination of mathematical morphology and Hough transform, the panchromatic band image of the original image of Resource-3 satellite is used as the experimental object to extract the road and evaluate the accuracy of the road extraction.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】