高分辨率影像道路提取研究

发布时间：2018-12-29 16:30

【摘要】：在遥感技术、智能计算和云计算计算高度发展的今天,虽然高分辨率遥感影像处理技术成果层出不穷,但高分辨率影像道路自动提取仍是当今研究热点之一。在诸如地图导航应用、土地资源监测、城乡规划等领域,高分辨率影像道路自动提取已经成为不可或缺又亟待提升关键技术。如何在高分辨率影像复杂的纹理和几何信息中准确的获取道路信息仍然是高分影像目标提取的重点课题。传统的影像道路提取,尚未有效地顾及高分辨率影像中丰富的纹理和几何特征,提取结果存在不同程度的问题。为此,本文提出了基于高分辨率影像道路多种特征进行综合分析的方法,进行高分辨率影像道路提取,其中采用正向定位和反向剔除的提取思想,利用最小外接矩形和最小外接圆,从多方面考虑道路提取条件,较好解决了高分辨率影像道路提取问题。本文主要研究高分辨率航空影像各类特征进行提取方法,获取道路的纹理特征和几何特征,采用正向定位和反向剔除的整体思路对道路进行定位。提出使用纹理和Gabor提取的方式确定道路的位置,采用最小外接矩形和最小外接圆对道路之外的地物进行搜索并剔除,从道路本身和道路之外的地物两个方面着手,配合影像分割方法、运用多特征分析、最小外接矩形和外接圆提取和数学形态学的应用准确的对道路进行提取。本文多特征综合分析提取的流程为:(1)影像道路纹理特征提取(2)影像道路区域分割(3)影像道路分类(4)多特征分析(5)数学形态学精细处理。最后利用C++设计开发了高分辨率影像道路提取实验平台,分别选取城市道路影像和山区道路影像为实验影像。实验表明,本文提出多特征综合分析处理方法能较好的提取高分辨率遥感影像中道路结构,可以有针对性的设定阈值对影像进行处理。相比较前人提取算法,在提取效果方面有较大提升。在实验过程中发现,城市道路提取和山区道路提取有一定的阈值设定规律,并有针对性的加以总结。虽然本文提出的流程和提取分析方法能够较好的提取道路,但是在阈值设定的自适应性方面还需要做进一步的研究,同时针对阴影遮挡较为严重的路面的提取也需要进一步深入研究。
[Abstract]:With the development of remote sensing technology, intelligent computing and cloud computing, although high-resolution remote sensing image processing technology results in endlessly, but high-resolution image road automatic extraction is still one of the hot spots. In such fields as map navigation, land resource monitoring, urban and rural planning, high-resolution image automatic road extraction has become an indispensable and urgent need to improve the key technology. How to obtain road information accurately in the complex texture and geometry information of high resolution image is still the key task of target extraction in high resolution image. Traditional road extraction has not yet taken into account the rich texture and geometric features of high-resolution images, and there are some problems in the extraction results. Therefore, this paper puts forward a comprehensive analysis method based on many features of high-resolution image road, and carries out road extraction of high-resolution image, in which the idea of forward location and reverse culling is adopted. By using the minimum outer rectangle and the minimum circumscribed circle, the road extraction condition is considered from many aspects, and the problem of road extraction in high resolution image is better solved. In this paper, we mainly study the methods of extracting all kinds of features from high-resolution aerial images, obtain the texture features and geometric features of the road, and use the whole idea of forward location and reverse culling to locate the road. The method of texture and Gabor extraction is used to determine the location of the road, and the minimum outer rectangle and the minimum circumscribed circle are used to search and eliminate the objects outside the road, which start from two aspects: the road itself and the objects outside the road. In combination with the image segmentation method, using multi-feature analysis, the minimum external rectangle and circumscribed circle extraction and mathematical morphology of the application of accurate road extraction. The process of multi-feature analysis and extraction is as follows: (1) road texture feature extraction (2) image road segmentation (3) image road classification (4) multi-feature analysis (5) mathematical morphology fine processing. Finally, an experimental platform for high resolution image road extraction is developed by using C, and urban road images and mountain road images are selected as experimental images. The experimental results show that the multi-feature comprehensive analysis and processing method proposed in this paper can extract the road structure in the high-resolution remote sensing image and can set the threshold to process the image. Compared with the previous algorithms, the extraction effect has been greatly improved. In the course of experiment, it is found that there are certain threshold setting rules in urban road extraction and mountain road extraction, and summarized accordingly. Although the process and extraction analysis method proposed in this paper can extract the road well, but in the adaptive aspect of threshold setting, we still need to do further research. At the same time, the extraction of more serious shading pavement also needs further research.
【学位授予单位】：北京建筑大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】