A Method for Solving Point Cloud Registration Models Using Dual Quaternion Descriptions of Point‑Planar Feature Constraints

Objectives The high-precision registration of point cloud data is the key to ensure the integrity of 3D data on the surface of spatial objects. To address the problem that there are differences in position, attitude and scale of cloud data from neighboring stations, a method is proposed to solve the registration model of point cloud described by the dual quaternion under the constraints of point-planar feature.

Methods First, the rotation matrix and translation vector of the spatially similar transformation are represented by the dual quaternion, based on which the scale factor is taken into account and the vertical and parallel spatial topological relationships exist between the vectors constructed by the points in the plane and the points out of the plane respectively and the normal vectors of the plane, and this is used as the constraint of the spatially similar transformation to construct the parity model based on the least squares criterion. Then the Levenberg-Marquardt method is introduced to solve the level-difference model to avoid the possible non-convergence of the iterations in the level-difference treatment due to the inappropriateness of the initial values or due to the fact that the real symmetric matrix constructed by the Jacobi matrix is close to singularity.

Result Two sets of experiments are compared and analyzed with the existing methods, and the experimental results show that the proposed method can effectively achieve point cloud registration.

Conclusions Therefore, the method that takes into account the scale factor under the point-planar feature constraint and uses the dual quaternion to realize the spatial similarity transformation has a strong practical value.