Surface Texture and Microstructure Extraction from Multiple Aerial Images

摘要

Textures and microstructures on the surfaces of cultural objects, such as buildings and parking lots, provide important information for many military and civilian applications. The extraction of this information from aerial imagery is difficult due to problems arising from perspective distortion, data deficiency, and corruption caused by shadows and occlusions. In this paper we present a generic surface texture and microstructure extraction system to deal with these problems, under the assumption that an initial 3-D geometric site model is given and sufficient camera and light source information is known. The infrastructure of the system is an orthographic facet image library that systematically collects surface intensities from multiple aerial images into a database, eliminates the effects of shadows and occlusions, and forms a complete and consistent surface intensity representation. A robust, local intensity based image analysis module is designed to symbolically extract surface microstructures with rectilinear and repetitive patterns. The paper shows successful applications of the system to extractions of both 2-D and 3-D microstructures. The system is typically useful for urban site model refinement and visualization.