Cameras are appealing sensors for airborne applications, since they measure passively and are small, lightweight, and inexpensive. Compared to other sensors used for navigation, they feature complementary error characteristics. They capture the environment, and hence provide observations for the solution of the problem commonly referred to as simultaneous localization and mapping. In photogrammetry, bundle adjustment is known to provide optimal estimates for simultaneous scene reconstruction and pose estimation. A sliding-window adjustment is presented that features real-time capability: that is, a continuous video stream can be processed by the usually limited onboard resources of an unmanned aerial vehicle. Having airborne applications in mind, a parametrization for unknown three-dimensional points is incorporated that is able to represent points located far away from the camera or even at infinity. Due to the delayed initialization of these points, the need for prior information about the depths of these points is avoided. The feasibility of these concepts is demonstrated by means of real data captured by a single geometrically calibrated camera mounted on an unmanned aerial vehicle. The performance of the approach is evaluated by comparing the achieved results with estimates obtained by a full bundle adjustment.
Sliding-Window Adjustment for Monocular Simultaneous Localization and Mapping
Journal of Aerospace Information Systems ; 12 , 12 ; 745-755
2015-07-24
11 pages
Article (Journal)
Electronic Resource
English
Research on Simultaneous Localization and Mapping Algorithm Based on Monocular Vision
British Library Conference Proceedings | 2018
|A robust approach for a filter-based monocular simultaneous localization and mapping (SLAM) system
BASE | 2013
|Simultaneous Localization and Mapping
Tema Archive | 2010
|Robocentric Simultaneous Localization and Mapping
AIAA | 2024
|Monocular Localization of a Mobile Robot
British Library Conference Proceedings | 1993
|