基于知识规则构建和形态学修复的建筑物提取研究  

A Study of Building Extraction Based on Morphological Rehabilitation and Rule-Oriented Classification

作　　者：黄金库[1,2] 冯险峰[1] 徐秀莉[3] 丁青[1,2] 

机构地区：[1]中国科学院地理科学与资源研究所,北京100101 [2]中国科学院研究生院,北京100049 [3]北京世纪安图数码科技发展有限公司,北京100085

出　　处：《地理与地理信息科学》2011年第4期28-31,F0002,共5页Geography and Geo-Information Science

基　　金：国家科技攻关项目(2008BAK50B06)

摘　　要：提出了一种基于知识规则构建和形态学修复的建筑物提取方法,并以青岛市崂山区作为试验区开展建筑物信息提取研究。首先,选用试验区的快鸟影像,进行影像预处理、小尺度分割、大尺度合并、知识规则构建等算法处理,得到建筑物轮廓的粗提取结果,对提取结果经过形态学修复和边缘检测后,得到粗提取的建筑物轮廓矢量图;然后在以ArcGIS Engine为平台开发的建筑物提取系统中,以预处理后影像和第一步中获取的建筑物矢量图作为双底图,针对建筑物的不同形态,分别采用手扶跟踪数字化、自动跟踪数字化和模型数字化来规则化处理建筑物轮廓;最后获得建筑物轮廓的精提取结果。试验结果表明,与监督分类方法相比,这种方法提取出的建筑物轮廓清晰完整、精度高、速度快,提高了建筑物提取的自动化和智能化水平。Building extraction is a significant studying realm of Remote Sensing.In this paper,a morphological rehabilitation and rule-oriented classification method is proposed,and Laoshan District,Qingdao City is used as the experimental area.Firstly,a coarse building contour is obtained from the Quick Bird image through the algorithms including pretreatment,small-scale segmentation,merging segmentation,rule-based classification,morphological rehabilitation and edge detection.Secondly,to make the building contour more accurately,an extraction system is developed with the ArcGIS Engine platform.According to the different morphology of building contours,methods,such as manual tracking digitization,automatic digitization and model digitization,are adopted in this system,respectively.At this stage,both the vector building contour obtained from the first step and the preprocessed remote sensing image are treated as basic maps.Finally,precise building contours are extracted.Compared with the supervised classification,the method can extract buildings more effectively,and lets the work of building extraction become much more automatic and intelligent.

关 键 词：高分辨率影像 建筑物提取 模型库 图像分割 数字化 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]