Optimal geometric model matching under full 3D perspective
Computer Vision and Image Understanding
Performance Evaluation and Analysis of Monocular Building Extraction From Aerial Imagery
IEEE Transactions on Pattern Analysis and Machine Intelligence
The ascender system: automated site modeling from multiple aerial images
Computer Vision and Image Understanding
Retrieving Shape Information from Multiple Images of a Specular Surface
IEEE Transactions on Pattern Analysis and Machine Intelligence
Building Reconstruction from Optical and Range Images
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
Acquisition of a Large Pose-Mosaic Dataset
CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
| Hi-index | 0.00 |
We introduce an algorithm that fuses information from aerial and terrestrial views for the automatic reconstruction of high-resolution building models within built-up areas. Calibrated aerial photography is commercially available for wide areas of coverage and has been shown to be a useful source of information about the location of buildings at the site, their 2D footprint [8,10], and their rooftop shape [1,6,9]. In contrast, terrestrial imagery is usually uncalibrated, not available commercially for most urban areas, and difficult to acquire. These ground-level images do, however, provide close-range, high-resolution views not normally available in aerial data. Our approach uses the pose information typically associated with aerial surveillance imagery to acquire an initial three-dimensional model of the buildings at the site. Uncontrolled, terrestrial imagery is then aligned to the model using a symbolic model matching and pose a refinement technique. Once aligned, ground-level views can be used to enhance the site model in a number of ways. High-resolution façade textures can be mapped onto the model geometry using the recovered pose information and standard texture mapping algorithms. The same algorithms allow explicit segmentation of building facades from terrestrial views as regions of pixels that project to vertical structures in the model. Context sensitive processing can be applied to these façade regions for the symbolic extraction of surface structures such as windows, doors, and pillars.