TEXTURE MAPPING FOR BUILDING FACADES USING TERRESTRIAL LIDAR POINT CLOUDS AND CLOSE RANGE IMAGES

摘要

Facade texture mapping for building models has been widely studied to meet the needs of photorealistic building models for more than fifteen years. Except from using texture library, spectral information from images is needed to realize photorealistic visualization. Close range imagery is a preferable material because it provides information of the building facade from human perspective. However, in many situations, textures would not be proper to directly drape due to image occlusions on a single building facade. With accurate depth information, terrestrial lidar data can be used to distinguish occlusions from target facade. On the other hand, images taken from different directions help each other to compensate the occluded parts. To strike a balance between reality and visualization, we assume that one facade image patch is able to substitute other facade patches provided that they have the same texture. Therefore, texture analysis is used to recognize and classify the facade textures. Experimental results show that the proposed method may reach high fidelity, provided that the detailed building models are available.