摘要
基于航空遥感图象的道路自动检测与识别是数字城市、道路监控等领域非常重要的课题;其检测方法经改进可用于跑道识别,为飞机自动降落提供必要的跑道信息。该文以传统道路提取方法为基础进行改进,提出一种道路自动提取方法,该方法包括以下几步:首先,利用微分几何方法对图象进行线特征强化处理;接着,进行线段提取,将传统方法中需要分步实现的线特征抽取和线段连接同步完成,同时对连接判定阈值的设置进行了一些讨论,提出了自动设置阈值的方法;最后,根据道路的拓扑性质进行线段的聚合分类。实验结果表明,该方法能准确有效地提取出道路。
Automatic road extraction from aerial imagery is a very important technology for construction of digital cities and inspection of road networks.And it can provide necessary information for air vehicles' automatic landing by some improvement.Based on the traditional one,an algorithm for automatic road extraction from aerial remote sensing images is presented,which consists of three processing levels.At first,the linear features in the original image are enhanced by utilizing the differential geometric characteristic of roads.Secondly,two first steps in traditional methods,that is,the extraction of linear features and connection of road seeds are synthesized as lines extraction with the local characteristic of image and some contextual information both used.In addition,the problem of thresholds setting is discussed,and an automatic method is proposed.At last,a perceptual grouping plan based on the topological information of roads is introduced.Experiments show the method can accurately extract roads.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第34期218-221,共4页
Computer Engineering and Applications
基金
航空科学基金(编号:01F53028)
关键词
道路提取
特征强化
线段提取
线段聚合
road extraction,feature enhancement,connecting technique,perceptual grouping
