高分辨率遥感影像建筑物轮廓提取方法研究

发布时间：2019-06-15 11:26

【摘要】：随着遥感技术不断发展,高分辨率遥感影像的获取更加简单快捷,高分辨率遥感影像的应用也越来越广泛。通过高分辨率遥感影像获取各类信息是高分辨率遥感影像的重要应用,也是人们获取各类地理信息的重要途径。对高分辨率遥感影像中的地物信息进行提取,并且快速的生成矢量地图对国民经济发展、生产生活需要以及军事等各个领域方面都有着重要的作用。建筑物作为典型的人工地物,对其进行信息提取有着十分重要的学术价值与应用前景。本文主要做了以下研究:(1)本文研究了前人使用高分辨率遥感影像进行建筑物轮廓提取的方法,在这些方法的基础上进行了分析与总结。针对高分辨率遥感影像建筑物呈现为面状结构、边缘特征明显以及房顶颜色丰富易辨的特性,提出了本文的基于边缘检测的建筑物轮廓提取方法和基于彩色影像分割的两种建筑物轮廓提取建筑物方法。(2)研究了基于边缘检测的建筑物轮廓提取方法。通过对传统的建筑物轮廓提取方法的研究,将轮廓提取分为了边缘检测以及轮廓提取两个部分。在边缘检测部分中,本文对已有的边缘检测算法进行分析归纳,重点对Sobel算子进行改进。将Sobel算子的模板由垂直和水平两方向扩展为四方向和八方向。然后,采用Otsu’s的方法选择一个最佳全局阈,对梯度计算的结果进行边缘点判断。在轮廓提取部分,本文对二值边缘影像进行直线检测,拟提取出建筑物轮廓信息,但是并未取得较好的成果,分析了其中的原因。(3)研究了基于彩色影像分割的建筑物轮廓提取方法。利用高分辨率遥感影像色彩信息丰富,在农村地区建筑物与周围地物之间存在较大的色彩差异这一特点,本文将高分辨率遥感影像分为R、G、B三个分量图像,根据不同的分量图中,具有相同或者相似性质的区域内像素值具有相同或者相似的平均或者期望的颜色估计,对高分辨率遥感影像进行影像分割,然后对分割完成后的区域进行边缘检测,接着通过Hough变换对边缘进行直线检测、平行线判断等一系列后续处理最终得到建筑物的轮廓。经过一系列的仿真测试表明,改进后的Sobel算子与改进前相比较能更加完整的反应出建筑物的边缘,并且能够更加有效的避开阴影对边缘检测带来的影响;基于彩色影像分割的建筑物提取方法能有效地提取出建筑物的轮廓线,并且能有效的减少“同谱异物、同物异谱”现象。
[Abstract]:With the continuous development of remote sensing technology, the acquisition of high-resolution remote sensing images is more simple and fast, and the application of high-resolution remote sensing images is becoming more and more extensive. Obtaining all kinds of information through high resolution remote sensing image is an important application of high resolution remote sensing image, and it is also an important way for people to obtain all kinds of geographical information. The extraction of ground object information from high resolution remote sensing images and the rapid generation of vector maps play an important role in the development of national economy, the needs of production and life, as well as the military and other fields. As a typical artificial object, it is of great academic value and application prospect to extract information from buildings. The main contents of this paper are as follows: (1) in this paper, the methods of building outline extraction using high resolution remote sensing images are studied, and on the basis of these methods, the methods are analyzed and summarized. In view of the fact that the building in high resolution remote sensing image is planar structure, the edge feature is obvious and the color of roof is rich and easy to distinguish, the building outline extraction method based on edge detection and two building outline extraction methods based on color image segmentation are proposed in this paper. (2) the building outline extraction method based on edge detection is studied. Through the research of the traditional building outline extraction method, the outline extraction is divided into two parts: edge detection and contour extraction. In the part of edge detection, the existing edge detection algorithms are analyzed and summarized, and the Sobel operator is improved. The template of Sobel operator is extended from vertical and horizontal to four and eight directions. Then, the Otsu's method is used to select an optimal global threshold, and the edge points of the gradient calculation results are judged. In the outline extraction part, this paper carries on the straight line detection to the binary edge image, intends to extract the building outline information, but has not obtained the good result, has analyzed the reason. (3) has studied the building outline extraction method based on the color image segmentation. Taking advantage of the rich color information of high resolution remote sensing image and the large color difference between buildings and surrounding objects in rural areas, the high resolution remote sensing image is divided into R, G, B component images. According to different component images, the pixel values in the region with the same or similar properties have the same or similar average or expected color estimation. The high resolution remote sensing image is segmented, then the edge of the segmented area is detected, and then the edge is detected by Hough transform, and the outline of the building is finally obtained by a series of subsequent processing, such as line detection, parallel line judgment and so on. A series of simulation tests show that the improved Sobel operator can reflect the edge of the building more completely than before, and can avoid the influence of shadow on edge detection more effectively. The building extraction method based on color image segmentation can effectively extract the contours of the building, and can effectively reduce the phenomenon of "homospectral foreign body, isomorphism".
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P237

【相似文献】