[PDF] Bundle Adjustment - A Modern Synthesis | Semantic Scholar (2024)

This paper proposes a new parameterization of a perspective camera based on quaternions, with no redundancy in dimensionality and no constraints on the rotations, and achieving the same performance in all investigated parameters.

Experimental results show that the computational cost and convergence speed is further improved to achieve similar accuracy without explicit adjustment over the structure parameters, and an implicit modeling approach based on a Direct Linear Transform (DLT) estimation is proposed.

Different strategies to overcome the quadratic computation time of Structure-from-Motion calibration algorithms by only working on subsets of images and merging the results are presented.

This paper introduces a fast and robust structure-less alternative to full bundle adjustment based on optimizing algebraic errors for trilinear constraints from triplets of views and shows good convergence properties with a large convergence basin and solutions that are close to the optimal solution.

This work proposes an improved, camera-orientation independent cost function that can be used instead of the standard bundle adjustment cost function, which yields a new bundle adjustment formulation which exhibits noticeably better numerical behavior, but at the expense of an increased computational cost.

This paper reports for the first time in the literature the complete derivation of the Jacobian matrix for bundle adjustment with constrained cameras, to foster other implementations of oblique multi-camera systems.

A novel approach to bundle adjustment for large-scale camera configurations by using epipolar and trifocal constraints to implicitly establish the relations between the cameras via the object structure and handling points far from the camera in a numerically stable fashion is proposed.

Results are demonstrated that suggest the ability of the algorithm to rival methods of RANSAC based pose estimation polished by bundle adjustment in terms of solution robustness, speed and accuracy, even when given poor initialisations.

...

...

It is shown that a simple pre-conditioning step removes the effect of the choice of coordinate frame, and together with a set of enforced constraints on the reconstruction, achieves along with this invariance greatly increased convergence speed over existing methods.

A numerical library for estimating multi-image matching constraints, or more precisely the multi-camera geometry underlying them, designed to be modular and open-ended, so that new feature types or error models, new constraint types or parametrizations, and new numerical resolution methods are relatively easy to add.

A bundle adjustment algorithm is formulated whose results are independent of both the coordinate frame chosen to represent the scene and the ordering of the images, which is more efficient that existing approaches to the problem in photogrammetry.

A program is described for carrying out least-squares camera modelling and scene reconstruction from a set of image and scene measurements of geometric features, easily extendible to include very general types of camera, image feature or measurement.

The main theoretical advance here is showing how to adjust the system information matrix when scene/camera parameters are removed from the reconstruction, and thus how to achieve an efficient recursive solution to the reconstruction problem.

This paper addresses the problem of determining the kind of three- dimensional reconstructions that can be obtained from a binocular stereo rig for which no three-dimensional metric calibration data is available, and shows that even in this case some very rich non-metric reconstructions of the environment can nonetheless be obtained.

This paper describes a new, more direct approach to geometric fitting, formulating it as the explicit recovery of a coherent, statistically optimal set of estimates of the “underlying data points” that gave rise to the observations, together with the estimated constraints which these points exactly verify.

By examining the eigenvectors associated with null or small eigenvalues of the Hessian matrix, this paper can quantify the exact nature of these ambiguities and predict how they will affect the accuracy of the reconstructed shape.

...

...