Accurate and True Reconstruction of 3D Models with Closed-Form Solutions
No Thumbnail Available
Date
1998-12-01T00:00:00Z
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
3D models of objects or scenes can be reconstructed from image sequences or multiple views. Reconstruction algorithms that adopt calibrated cameras generally produce more accurate results than those that adopt uncalibrated cameras, whereas the latter are more convenient to use. Depending on an application's need, one may trade off between reconstruction accuracy and convenience.
The objective of this research is to devise an accurate method for reconstructing the true and complete 3D models of small objects for use in applications such as virtual reality and model-based object recognition. This technical report focuses on the task of accurately recovering 3D coordinates of world points given a set of tracked 2D image points.
To ensure reconstruction accuracy, the method uses a calibrated camera and a computer-controlled turntable. With this equipment setup, close-form solutions to the reconstruction problem are derived for various camera models including the perspective and the projective models. The existence of close-form solutions facilitate the construction of an efficient, accurate, and reliable optimization algorithm that performs true and complete reconstruction.
Comprehensive experiments have been conducted to test the method's performance. Experimental results show that the method is more accurate and robust against noise than existing methods.