Automatic Model Reconstruction of Indoor Manhattan-World Scenes from Dense Laser Range Data
- Angela Budroni
Nowadays image-based modeling is receiving much attention and many applications require precise and photo-realistic 3D models. The camera calibration and orientation phases are key steps in the 3D modeling process. If these phases are not accurately performed, there will be some errors in the final model and for some applications low accuracy results are not accepted. The goal of this work is to investigate the influence of wrong camera parameters or bad image configuration in object reconstruction. The analysis is performed with a bundle adjustment solution perturbing the interior camera parameters and using different network configurations. We analyze the effects of wrong focal length and principal point as well as absent distortion parameters with images acquired under typical project configurations. Finally we report some examples of 3D modeling of complex architectures where the theoretical considerations cannot always be fulfilled.