Robust Bundle Adjustment for Structure from Motion

Abstract

Structure from motion (SFM) is the problem of reconstructing the geometry of a scene from a stream of images. In this problem, the geometry of the scene must be inferred from images, along with the camera pose parameters. Bundle Adjustment (BA) is a refinement method used to improve SFM solutions. It consists in simultaneously improving a set of initial estimates for all parameters (structure and camera pose) by minimizing a global cost function. It is generally considered to be highly accurate, and so is typically used as a last refinement step in most current SFM methods. Unfortunately, estimating the pose of the camera from a stream of images is an ill-conditioned problem. We thus propose a BA adjustment formulation which does not involve solving for the camera orientations. We tested this approach on several real world models. The numerical results obtained show that this approach is much less affected by noise than traditional BA.

Angle Independent Bundle Adjustment Refinement

“…In earlier work, we developed a related technique for improving bundle adjustment using a degree-four polynomial formulation [13]. In that method, while fewer unknowns need to be optimized, the expressions do not produce a single consistent set of both scene points and camera centers (as with typical bundle adjustment methods) and the computations are not automatically partitioned into sub-problems and thus cannot handle large datasets.…”

Extracting three-dimensional (3D) spatial information from sequential oblique unmanned aerial system (UAS) imagery for digital surface modeling

“…Consequently, the stem of SfM can be calculated by setting up an initial stereo model and incrementally adding additional images. Lastly, the error propagation is optimized via bundle adjustment, minimizing the projection error towards all spatial information obtained (Börlin et al 2019;Grün 1982;Kanatani and Sugaya 2011;Kyung, Kim, and Hong 1996;Triggs et al 2000;Zhang, Boutin, and Aliaga 2006).…”

Modelling of 3D Objects Using Unconstrained and Uncalibrated Images Taken with a Handheld Camera

“…Each image is initialised with the same rotation and focal length as the image to which it best matches. This has proved to work very well even though images have different rotation and scale [26,24,4]. Figure 4 (a) shows one image of a set of images of a paper model of the Daliborka tower at Prague castle [3].…”