Toward a system for road network automatic extraction in land consolidation using high spatial resolution imagery

Land consolidation is a tool for increasing the area of the arable land and improving the effectiveness of land cultivation. This paper presents a practical system for automatic road extraction in land consolidation to monitor the implementation of the project. The system integrates processing of color image data and information from digital spatial databases, takes into account context information, employs existing knowledge including plans of land consolidation, rules and models, and treats each road subclass accordingly. The system was designed as three-tier construction including interface, modules and database. The prototype system has been implemented as a stand-alone software package, and has been tested on a large number of images in different land consolidation areas. The parallel line segments are firstly detected and then the improved Active Contour Models (Snakes) are introduced to link the extracted road segments to the whole networks. The system was pilot used in the study area of Fangshan in Beijing which achieved satisfactory results. Key-Words: system, road extraction, land consolidation, high spatial resolution imagery, Snakes