摘要
提出了一种新的从航空城区图像中自动提取矩形建筑物的方法 .该方法基于从航空城区图像中提取的边缘 ,经过轮廓跟踪 ,采用Splitting方法提取直线 ,得出其相应的直线几何图形 ;针对航空图像的复杂及现有边缘检测算法的不足 ,提出了一系列直线处理的方法 (如直线的分类、排序、合并、调整等 )有效地弥补了前述处理的不足 ;为提高矩形房屋提取的准确率 ,引入知识定义了几种近似的矩形结构 .文章采用几何结构元分析的方法 ,提取图形中构成矩形的各种基本结构元 ,再根据结构元合并的准则 ,将各种基本结构元通过一定的合并算法合并成矩形结构 .大量试验结果证明该方法提取矩形房屋的准确率较高 ,鲁棒性好 ,运算速度快 。
This paper describes the development of a new Primitive Structure based approach for rectangle building, extracted from aerial imagery. The approach concentrates on two aspects: to systemize the approach and to develop robust and fast algorithms. The process of building detection is divided into four small stages, namely line extraction, line processing, and rectangle primitive structure searching, rectangle structure confirming and rebuilding. Robustness is achieved by only considering the geometric relations between lines in the course of building extraction. Edge elements are detected by a Canny edge filter, while linear elements are extracted using splitting method. This paper develops several line processing method including classifying and ordering the lines, adjusting the angle of the lines, merging the lines, and deleting the short lines to offset the defects of the previous processing methods. Several rectangle primitive structures are defined, according to the human knowledge, to enhance the probability of building detection. Firstly we first obtain all kinds of primitive structure that comprise rectangle structure from the straight line graph. Secondly merge these rectangle primitive structures to form rectangle buildings through the primitive structure merging algorithm. Finally, we can confirm and rebuild intact rectangle buildings. The experimental results show that the approach is computationally much faster than the previous methods. The results of perfomance avaluation also support the robustness and precision of the approach developed.
出处
《计算机学报》
EI
CSCD
北大核心
2003年第7期866-873,共8页
Chinese Journal of Computers
基金
基金项目 ( 4 13 0 10 70 1
5 14 0 10 2 0 2 0 1JW 0 5 )资助
