Advertisement
Published in:
2016 | OriginalPaper | Chapter
11. Bundle Adjustment
Authors : Kenichi Kanatani, Yasuyuki Sugaya, Yasushi Kanazawa
Published in: Guide to 3D Vision Computation
Publisher: Springer International Publishing
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Abstract
In the preceding chapter, we used knowledge of the camera matrices for reconstructing the 3D from multiple views. In this chapter, we describe a procedure, called bundle adjustment, for computing from multiple views not only the 3D shape but also the positions, orientations, and intrinsic parameters of all the cameras simultaneously. Specifically, starting from given initial values for all the unknowns, we iteratively update them such that the reconstructed shape and observed images better satisfy the perspective projection relationship. Mathematically, this is nothing but minimization of a multivariable function, but the number of unknowns is very large. Hence, we need to devise an ingenious scheme for efficiently storing intermediate values and avoiding unnecessary computations. Here, we describe a typical programming technique for such implementation.