基于高分辨率遥感影像的城区建筑物提取方法研究

发布时间：2018-04-25 13:40

本文选题：建筑物提取 + 形态学滤波　；参考：《吉林大学》2017年硕士论文

【摘要】：伴随着科技与遥感影像技术的不断发展与进步,遥感卫星种类不断增加,而性能也在不断的提升当中。很多遥感影像的分辨率达到了亚米级的水平,这为地物信息的研究提供了更加丰富的细节。我国是一个地大物博的国家,拥有着丰富的地理信息,伴随着国家的发展与城市的建设所产生的地貌变化也愈发巨大。所以能否有效的利用遥感影像所提供的信息对未来的城市建设与规划都有重要的意义。对于城市地区而言,大部分的地貌都是由建筑物覆盖,对于城区建筑物的位置与状态等信息的提取,能够给城市人口规划、管道布置、城市建设等民生问题提供重要参考。因此本文基于课题的实际需求,经过深入分析与研究,给出了一个准确、快速、高效的城区建筑物提取方法。针对于建筑物像素特点的一些方法被广泛应用于传统的提取方法中,普遍存在的问题是提取效果不理想,噪声偏多等。为了准确地提取出城市建筑物,本文设计出一种结合面向对象与最近邻分类的建筑物分级提取策略。首先,对于同质区域的提取采用了面向对象的思想来进行图像的分割,提出了一种改进的分水岭算法。然后根据形状性约束条件率先提取出分割完好的建筑物部分,并依据建筑物的普遍面积大小来对提取出的面积过小或者过大的建筑物信息进行选择性的过滤。再应用多方向性形态学滤波算法将那些与建筑物相邻且具有较高光谱特征相似性的道路剔除。最终根据之前得到的建筑物信息与分离出的非建筑物信息进行基于最近邻分类的进一步提取工作,从而使图像中所有的建筑物得到完整的提取。通过对实验结果的分析表明,对于高分影像中建筑物的提取,本文所采用的方法可以在一定的时间范围内给出一个准确清晰的提取结果。进而可以对城市的管理与规划做出贡献,同时也可以为市政管理部门提供强大的技术支持。
[Abstract]:With the development of science and technology and remote sensing image technology, the variety of remote sensing satellite is increasing, and the performance of remote sensing satellite is improving. The resolution of many remote sensing images reaches the sub-meter level, which provides more details for the study of feature information. China is a large country with abundant geographical information. With the development of the country and the construction of the city, the geomorphological changes are becoming more and more huge. Therefore, it is of great significance to use the information provided by remote sensing images for the future urban construction and planning. For urban areas, most of the landforms are covered by buildings. The extraction of information such as location and state of urban buildings can provide an important reference for urban population planning, pipeline layout, urban construction and other livelihood issues. Therefore, based on the actual demand of the subject, this paper presents an accurate, fast and efficient method for extracting urban buildings through in-depth analysis and research. Some methods based on the characteristics of building pixels have been widely used in traditional extraction methods. The common problems are that the extraction effect is not ideal and the noise is too much. In order to extract urban buildings accurately, this paper designs a hierarchical extraction strategy combining object-oriented classification with nearest neighbor classification. Firstly, an improved watershed algorithm is proposed for homogeneous region extraction by using the object oriented approach to image segmentation. Then according to the shape constraints, the partitioned part of the building is firstly extracted, and the information of the building which is too small or too large is selectively filtered according to the size of the general area of the building. Then the multi-directional morphological filtering algorithm is applied to remove the roads adjacent to the buildings with high spectral characteristics. Finally, according to the building information obtained before and the separated non-building information, further extraction based on the nearest neighbor classification is carried out, so that all buildings in the image can be extracted completely. Through the analysis of the experimental results, it is shown that the method used in this paper can give an accurate and clear result in a certain time range for the extraction of buildings in high score images. It can also contribute to the management and planning of the city and provide strong technical support for the municipal administration.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP751

【相似文献】