高分辨率遥感影像城市典型目标提取及街区功能分类方法研究

发布时间：2018-03-17 04:04

本文选题：高分辨率　切入点：遥感影像　出处：《山东农业大学》2017年硕士论文　论文类型：学位论文

【摘要】：工业革命使城市化得以在世界范围内快速的发展。可是在一些非综合功能分区的城市布局理念指导下,城市的最基本功能被人为强制性的割裂开来,相互间缺乏有机联系,使内部功能失去活性,给居民生活带来不便。街区作为城市的基本组成单元,是城市经济、文化、政治等活动的基础,其功能规划优劣直接影响着城市的建设与发展。因此,可以通过研究城市街区模式来研究城市模式,发现城市内部问题。可是由于城市街区信息量大、范围广等特点,导致获取城市街区信息需要耗费大量人力、物力,且周期长、效率较差。随着电子技术、光学成像技术、网络传输技术等的飞速发展,遥感对地观测技术具备了多角度、高空间分辨率、高时间分辨率、高光谱分辨率等优势,可以获取城市地物详细信息,且信息覆盖周期短。基于以上分析,本文提出高分辨率遥感影像城市典型目标提取及街区功能分类方法,其研究结果包含以下几个方面:(1)提出了一种面向对象的高分辨率遥感影像阴影提取方法。首先,利用mean shift算法进行地物特征聚类,去除噪声。然后,使用本文提出的阴影检测指数进行阴影检测。最后,利用阈值分割提取阴影区域。选取两景不同场景的实验数据进行了验证实验。实验结果表明,本文方法能准确、有效地提取阴影区域,且能去除水体、蓝色地物等非阴影地物的影响;另外,使用面向对象的思想可以有效地去除噪声的影响,提高检测的精确度。(2)提出一种基于感知编组的高分辨率遥感影像主干道路自动提取方法。首先,利用直线段检测器算法提取影像中直线段信息。然后,利用道路在高分辨率遥感影像上的几何特征进行感知编组。最后,经过长度约束得到道路信息。使用两景不同场景、不同传感器的实验数据进行验证实验。实验结果表明,两个实验中道路提取的完整率、正确率和检测质量都在96%以上。(3)提出一种高分辨率遥感影像城市街区功能分类方法。首先,通过计算获取建筑物、植被、水体、阴影的特征影像。然后,结合道路网信息将城市划分为独立的街区影像单元集合,并通过计算每个街区各特征影像均值的方式,将面向像元的处理方式转变为面向街区对象的处理方式。最后,通过LIBSVM分类器实现城市街区功能的分类。实验结果表明,本文提出的高分辨率遥感影像城市街区功能分类方法能将城中村、现代居民区、商业区等8类不同功能的街区很好的分类,且精度都在84%以上。
[Abstract]:The industrial revolution allowed the rapid development of urbanization around the world. However, under the guidance of the concept of urban layout of some non-comprehensive functional zones, the most basic functions of cities were cut apart by artificial compulsion, and there was a lack of organic connection between them. The block, as the basic component unit of the city, is the basis of the city's economic, cultural and political activities, and its function planning directly affects the construction and development of the city. We can study the urban model by studying the urban block model and find out the problems within the city. However, because of the characteristics of large amount of information and wide range of urban blocks, it takes a lot of manpower, material resources and a long period to obtain the information of urban blocks. With the rapid development of electronic technology, optical imaging technology and network transmission technology, remote sensing Earth observation technology has the advantages of multi-angle, high spatial resolution, high time resolution, high spectral resolution, etc. The detailed information of urban features can be obtained, and the information coverage period is short. Based on the above analysis, this paper puts forward a method of extracting typical urban targets and classifying the function of blocks in high-resolution remote sensing images. The research results include the following aspects: 1) an object oriented shadow extraction method for high resolution remote sensing images is proposed. Firstly, the feature clustering of ground objects is carried out by using mean shift algorithm to remove noise. The shadow detection index proposed in this paper is used for shadow detection. Finally, the shadow region is extracted by threshold segmentation. The experimental data of two different scenes are selected for verification. The experimental results show that the proposed method is accurate. The shadow area can be extracted effectively, and the influence of non-shadow objects such as water body, blue ground object and so on can be removed. In addition, the effect of noise can be effectively removed by using object-oriented thought. An automatic trunk road extraction method based on perceptual marshalling is proposed. Firstly, line segment detector algorithm is used to extract the line segment information in the image. Finally, the road information is obtained by length constraint. The experimental data of two different scenes and different sensors are used to validate the experiment. The experimental results show that, In the two experiments, the integrity rate, correct rate and detection quality of road extraction are above 96%.) A high resolution remote sensing image of urban block function classification method is proposed. Firstly, the buildings, vegetation, water body are obtained by calculation. Then, combining the road network information, the city is divided into a set of independent block image units, and by calculating the average value of each feature image in each block, The pixel oriented processing method is transformed into the block oriented object processing mode. Finally, the LIBSVM classifier is used to realize the classification of the urban block function. The experimental results show that, The high resolution remote sensing image of urban block function classification method proposed in this paper can classify 8 different function blocks, such as village in city, modern residential area, commercial district and so on, and the accuracy is more than 84%.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】