Pose Estimation from Corresponding Point Data

Abstract

Abstracr-Solutions for four different pose estimation problems are presented. Closed form least-squares solutions are given to the over constrained ZD-ZD and 3-D-3-D pose estimation problems. A globally convergent iterative technique is given for the 2-D perspective projection -3-D pose estimation problem. A simplified linear solution and a robust solution to the 2-D perspective projection-ZD perspective projection pose estimation problem are also given. Simulation experiments consisting of millions of hials having varying numbers of pairs of corresponding points, varying signal to noise ratio (SNR) with either Gaussian or uniform noise provide data suggesting that accurate inference of rotation and translation with noisy data may require corresponding point data sets having hundreds of corresponding point pairs when the SNR is less than 40 dB. The experiment results also show that robust technique can suppress the effect of blunder data that come from outliers or mismatched points.

Robust stereo tracking for space applications

“…The redundancy yields a more accurate result with the computation of the pseudo-inverse of L as given in equation (7). Furthermore if the number or the nature of visual features is modified over time, the interaction matrix L and the vector error s is easily modified consequently.…”

“…This study focuses on the registration techniques that allow alignment of real and virtual worlds using images acquired in real-time by moving cameras. In the related computer vision literature geometric primitives considered for the estimation are often points [7], [3], [11], contours or points on the contours [10], [2], [5], segments, straight lines, conics, cylindrical objects, or a combination of these different features [13]. Another important issue is the registration problem.…”

“…Linear approaches use a least-squares method to estimate the pose. Full-scale non-linear optimisation techniques (e.g., [7], [10], [5], [2], [14]) consist of minimising the error between the observation and the forward-projection of the model. In this case, minimisation is handled using numerical iterative algorithms such as Newton-Raphson or LevenbergMarquardt.…”

Registration of Video Images to Tomographic Images by Optimising Mutual Information Using Texture Mapping

“…Pose estimation can be achieved through corresponding pairs of 2D and 3D features, such as points [8] or points and lines [11]. The accuracy of the pose estimation algorithm is determined by the number of features, the accuracy with which the 2D and 3D coordinates of the features are known and the accuracy of their correspondence.…”

Smooth point-set registration using neighboring constraints

“…To cite a few, Dryden and Mardia (1998); Kendall (1984) deal with isometries and similarity transformations; Berge (2006); Umeyama (1991) deal with Euclidean transformations (i.e. excluding reflections from isometries); Haralick et al (1989) deals with similarity and projective transformations; and Hartley and Zisserman (2000) deals exclusively with projective transformations.…”