Out-of-Core Bundle Adjustment for Large-Scale 3D ReconstructionReport as inadecuate

Out-of-Core Bundle Adjustment for Large-Scale 3D Reconstruction

Out-of-Core Bundle Adjustment for Large-Scale 3D Reconstruction - Download this document for free, or read online. Document in PDF available to download.

Large-scale 3D reconstruction has recently receivedmuch attention from the computer vision community. Bundleadjustment is a key component of 3D reconstructionproblems. However, traditional bundle adjustment algorithmsrequire a considerable amount of memory and computationalresources. In this paper, we present an extremelyefficient, inherently out-of-core bundle adjustmentalgorithm. We decouple the original problem into severalsubmaps that have their own local coordinate systems andcan be optimized in parallel. A key contribution to ouralgorithm is making as much progress towards optimizingthe global non-linear cost function as possible using thefragments of the reconstruction that are currently in corememory. This allows us to converge with very few globalsweeps often only two through the entire reconstruction.We present experimental results on large-scale 3D reconstructiondatasets, both synthetic and real.

Computational Perception and Robotics - Computational Perception and Robotics Publications -

Author: Ni, Kai - Steedly, Drew - Dellaert, Frank - -

Source: https://smartech.gatech.edu/

Related documents