Research Of Automatic Registration And Fusion Technology With Multi-angle3D Data For3D Scanning System

Recent years, computer vision technology has attended people’s attention. And thethree-dimensional reconstruction studied in this paper is much important in the area ofcomputer vision. When we use the structured light scanning method to do the3Dreconstruction, we should scan the model from several angles and join these multiple piecesof data together. Therefore, the automatic registration of3D data is becoming a hot spot whichscientific research personnel has always been committed to. Calibration depending on theinstrument and automatic registration are the two data splicing methods which are often usedrecent years. For the small and medium models, depending on the precise turntable to splicethe point clouds from several angles is time-saving, simple and high precise. If we want tosplice the model precisely depending on the turntable, we should calibrate the axis of theturntable at first. Above all, we combine the two registration methods according to the actualsituation in this paper.However, depending on the single axis turntable to scan the model appears anotherproblem. The top and bottom part of model can not splice to the main part of model easily,because their overlap area is small and their initial position is uncertain. Using the ICPalgorithm to do the registration can not get a satisfactory result.This paper reports an automated multi-view three-dimensional data registration schemefor fully-automated three-dimensional scanning system. We use single-axis precise turntableto help to get multi-view point cloud and to integrate them together. First, we use a quick andeasy way to get the three-dimensional information of the turntable’s central axis, and thenintegrate the model body using the axis. After that we complete the initial registration of pointcloud based on minimum bounding box. At last we complete the precise registration usingimproved ICP algorithm.