Abstract: Automatic road extraction is the key factor of building map based on satellite image.However，due to the complexity of satellite image，the extraction results are not satisfying enough.Parallel pairs’ model conduces to the development of the research.An algorithm based on parallel pairs’ model and bottom-up control strategy is proposed.The algorithm consists of a low-level process for edge detection and linking，a mid-level process for the formation of road structure，and a high-level for process for the road recognition.In addition，necessary algorithms are also provided.Experimental results show that the algorithm can detect roads accurately and efficiently.

Key words: satellite image, roads extraction, edge detection, parallel pairs

摘要： 卫星照片中道路的自动提取是研究利用卫星照片构建地图的关键技术。卫星照片的复杂性往往导致提取的效果不理想。平行线对的提出有助于促进该研究的发展。通过使用平行线对模型，使用自下而上的控制策略，对卫星照片进行低层次的边缘检测和连接、中层次的道路特征结构化以及高层次的道路识别，并给出了相应的算法。实验结果表明，所提算法能够准确有效地提取卫星照片中的道路信息。

关键词: 卫星照片, 道路提取, 边缘检测, 平行线对