Incremental Light Bundle Adjustment for Structure From Motion and RoboticsReport as inadecuate


Incremental Light Bundle Adjustment for Structure From Motion and Robotics


Incremental Light Bundle Adjustment for Structure From Motion and Robotics - Download this document for free, or read online. Document in PDF available to download.

Bundle adjustment BA is essential in many robotics and structure-from-motion applications. In robotics, often a bundle adjustment solutionis desired to be available incrementally as new poses and 3D points areobserved. Similarly in batch structure from motion, cameras are typicallyadded incrementally to allow good initializations. Current incrementalBA methods quickly become computationally expensive as more cameraposes and 3D points are added into the optimization. In this paper weintroduce incremental light bundle adjustment iLBA, an efficient optimization framework that substantially reduces computational complexity compared to incremental bundle adjustment. First, the number of variablesin the optimization is reduced by algebraic elimination of observed3D points, leading to a structureless BA. The resulting cost function is formulatedin terms of three-view constraints instead of re-projection errorsand only the camera poses are optimized. Second, the optimization problemis represented using graphical models and incremental inference is applied,updating the solution using adaptive partial calculations each time anew camera is incorporated into the optimization. Typically, only a smallfraction of the camera poses are recalculated in each optimization step.The 3D points, although not explicitly optimized, can be reconstructedbased on the optimized camera poses at any time. We study probabilisticand computational aspects of iLBA and compare its accuracy against incrementalBA and another recent structureless method using real-imageryand synthetic datasets. Results indicate iLBA is 2-10 times faster thanincremental BA, depending on number of image observations per frame.



Computational Perception and Robotics - Computational Perception and Robotics Publications -



Author: Indelman, Vadim - Roberts, Richard - Dellaert, Frank - -

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







Related documents