Abstract: As a kind of base geographical information, road extraction from satellite images is important in the practice and theory of automatic satellite images recognition, and its results can be used for the data updating of geographical data base. Using existing approaches to automatic road extraction from satellite images for reference and according to the road properties in China, this paper presents a new road extraction method that combines image processing techniques with artificial intelligence methodology. In this approach, an operator is applied to an image to enhance the road information first, then road detection based on the shape index and other priori knowledge is performed to find possible road points. After using more contextual information or global constraints to extend the road seeds to form a road segment, the existence of a road is inferred and the gap between the road segments is connected by employing knowledge-based rules. Finally, experimental results on SPOT data are elaborated.

Key words: road extraction, satellite images, expert system, geographical information system

摘要： 公路信息是卫星遥感影像中的一种重要的地理专题信息，公路信息提取不仅在卫星数字图像自动解译方面具有理论与方法意义，而且可以利用提取的公路信息作为数据源，对地理信息系统数据库进行自动更新。针对我国公路的特点，提出了图像处理技术与人工智能相结合的新方法，以SPOT数字图像为数据源，对公路信息提取进行了研究。研究由卫星数字图像增强处理入手，通过突出目标与背景的边界差异，达到突现公路信息的目的，其次通过引入形状系数等先验知识，进行公路信息检测，找出可能的公路点。在此基础上，利用相关信息或全局约束来扩展公路种子点以形成公路片断，并通过基于知识的制导，利用各种知识来推断公路的存在并指导公路片断的连接。文章最后针对具体试验给出有关图像处理结果并进行了具体的讨论。

关键词: 公路信息提取, 卫星遥感图像, 专家系统, 地理信息系统