Self-alignment in marking edges of man-made objects

Abstract

The extraction of geometries of man-made objects, such as buildings and roads, from the images of air-borne sensors is the basis for digital city reconstruction. It has enormous benefits to the development of our society. Google Earth, for instance, provides details of cities around the world in 2D and 3D forms with the availability of precise information of man-made structures for anyone with Internet connection. Obtaining such geometric data using air-borne sensors, either semi-or fully automatically, is a very challenging task that inspire much research efforts in recent years. Fully automatic algorithms for extracting geometries of man-made objects is most desirable for real world applications. Due to the limitations of data and algorithms available, the current practical solutions still involve manual operators in key areas such as curvilinear feature identification and extraction. This paper presents some computerized aids to manual operations, such as autom atic alignment along the prominent structural edges on satellite images. These aids can effectively improve the efficiency and accuracy of manual operations. Various approaches are examined by their performances and potential for fully automatic deployment.