Cayley Transform of Position and Orientation Measurement with Monocular Vision

Abstract

In monocular vision-based position and attitude measurement, the non-linear iteration method based on feature-point geometric constraints is usually used to obtain results with higher accuracy. However, due to its strict geometric constraints to the feature points of target object and the requirement for highly-accurate initial value by iteration process, this algorithm is relatively poor in reliability. In this paper, Cayley transform is applied to the monocular vision-based position and attitude measurement; through an equivalent transformation from the unit orthogonal rotation matrix to 3D vectors by using Cayley transform, the computational process can be linearized, such that the complicated iterations caused by non-linear feature-point geometric constraints are eliminated, and the number of required feature points is reduced. Test data show that Cayley transform is practical and effective in the monocular vision-based position and attitude measurement.

DOI : http://dx.doi.org/10.11591/telkomnika.v12i3.4425