Photo-realistic 3D mapping from aerial oblique imagery

The generation of photo-realistic city model is an important task for many geospatial applications, such as 3D GIS, urban planning, virtual city tourism, car navigation, real estate, etc. The purpose of this research is to automatically generate façade texture from aerial oblique imagery. The developed algorithm can be applied to any oblique imagery taken from the air, such as UAV, airplane, or helicopter, as long as they were geo-referenced. Several problems will be encountered based on the use of oblique imagery. The major one is the occlusion comes from the building itself and the other buildings. This is serious especially for a congested urban area or a city with high-rise buildings. The proposed algorithm includes two major steps. The fist step is to generate all façade's texture within one image. However, the façade texture will be generated only when a certain geometric quality has achieved. The imaging geometry was estimated based on the incident angle between the camera-to-facade vector and the façade's normal direction. The second step is to detect and remove the occluded area based on existing 3D building models. Those two steps will be performed for all oblique images. However, in case one façade texture has been generated before and any marked occluded area has found, then a seamless Mosaicking procedure will be applied to achieve a complete texture. Meanwhile, due to errors always exist in exterior/interior orientation parameters and building models, some defect will occur along the boundary of occluded area. Thus, an amendment is necessary to remove this defect. Experimental results demonstrate that the proposed algorithm is effective and useful for automatic façade texture mapping.