On Cayley’s Factorization with an Application to the Orthonormalization of Noisy Rotation Matrices

A real orthogonal matrix representing a rotation in four dimensions can be decomposed into the commutative product of a left- and a right-isoclinic rotation matrix. This operation, known as Cayley’s factorization, directly provides the double quaternion representation of rotations in four dimensions...

Descripción completa

Detalles Bibliográficos
Autores: Sarabandi, Soheil, Perez-Gracia, Alba, Thomas, Federico
Tipo de recurso: artículo
Estado:Versión aceptada para publicación
Fecha de publicación:2019
País:España
Institución:Consejo Superior de Investigaciones Científicas (CSIC)
Repositorio:DIGITAL.CSIC. Repositorio Institucional del CSIC
OAI Identifier:oai:digital.csic.es:10261/202264
Acceso en línea:http://hdl.handle.net/10261/202264
Access Level:acceso abierto
Palabra clave:Rotation matrices
Quaternions
Double quaternions
Cayley’s factorization
Shepperd-Markley method
Descripción
Sumario:A real orthogonal matrix representing a rotation in four dimensions can be decomposed into the commutative product of a left- and a right-isoclinic rotation matrix. This operation, known as Cayley’s factorization, directly provides the double quaternion representation of rotations in four dimensions. This factorization can be performed without divisions, thus avoiding the common numerical issues attributed to the computation of quaternions from rotation matrices. In this paper, it is shown how this result is particularly useful, when particularized to three dimensions, to re-orthonormalize a noisy rotation matrix by converting it to quaternion form and then obtaining back the corresponding proper rotation matrix. This re-orthonormalization method is commonly implemented using the Shepperd–Markley method, but the method derived here is shown to outperform it by returning results closer to those obtained using the singular value decomposition which are known to be optimal in terms of the Frobenius norm.