The Effect of Point Density on the Registration Accuracy of a Terrestrial Laser Scanning Dataset

摘要

The terrestrial laser scanning method is widely used for creating three-dimensional (3-D) digital models of objects or scenes. One critical step in processing of terrestrial laser scanner measurements is the registration of point clouds into a common coordinate system. Thus, it is an important research topic for geosciences community. The iterative closest point (ICP) is mostly preferred to registration of point clouds in 3-D modelling studies. It performs the registration iteratively by the closest corresponding points. In this study, the effect of point density on registration accuracy of point clouds by the ICP was investigated. For this purpose, the laser scans were conducted with different point density, and the registration was executed by the ICP. Then the accuracies of the registrations were compared by the residuals of the check point coordinates. This study shows that the low point density slightly reduces the registration accuracy of the ICP. Nevertheless, since the registration parameters are computed using the high number of points, the ICP satisfies the expected accuracy for the work.