期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2007
卷号：XXXVI-4/W54
页码：127-134
出版社：Copernicus Publications
摘要：The goal of this work is to create a system for improving large architectures models with relatively few uncalibrated image sequences and produces interactively navigable and photorealistic 3D scene. The novel idea is based on the combination of spatial 3D vector data and uncalibrated image sequences to improve the fidelity of 3D architecture models in an extended urban area. First, the spatial 3D vector data is stored in our designed spatial 3D object-relational database. 3D vector data can simplifies the 3D reconstruction problem by supporting and solving directly for the architectural dimensions of the scene. Next, an architecture building is observed from multiple viewpoints by freely moving a video camera around it for taking uncalibrated images. All viewpoints are then linked by controlled correspondence linking for each image pixel. The correspondence linking algorithm allows for accurate depth estimation. Dense surface reconstructions are obtained by adjacent images of the sequence as stereoscopic pairs and computing dense disparity maps for each image pair. A flat wall of the architecture can be computed as a relief. Finally, we displace the surfaces of the model to make them maximally consistent with their appearance across these multiple images. Result shows that it is possible to compute a more realistic 3D architecture model using relatively few image sequences, as long as appropriate 3D vector data is available
