A Linear Features-Constrained, Plücker Coordinates-Based, Closed-Form Registration Approach to Terrestrial LiDAR Point Clouds

Considering that the low accuracy of extracted point features may affect the seamless fusion of point clouds from two neighbor stations, and by using traditional iterative-form solutions to implement point clouds registration, the large amount of computer resources, the high dependence on initial values of unknown parameters, and its theoretical instability in solving transformation parameters for large-angle registration can hardly be neglected. To alleviate the above problems, a linear features-based, closed-form solution to registration of pairwise terrestrial LiDAR point clouds is proposed, in which Plücker coordinates is introduced to represent linear features in 3D space. A Plücker coordinate-based object function is first introduced on the assumption of the consistency of each conjugate linear features from the two neighbor stations after registration. Based on the theory of least squares and by extreme value analysis of the error norm, detailed derivations of the model and the main formulas are all given. Experiments show that the proposed algorithm is just the one expected, the linearization of multivariate function is neglected in the implementation, and it runs well without initial estimates of unknown parameters, which assures the stability in solving transformation parameters for large-angle registration problems. Furthermore, by employing linear features as registration primitives, random errors may be greatly decreased by fitting contrast to point features based registration algorithms.