Automation of scan process for shape determination - using robot arm and triangulation scanner

Abstract

In recent years, the rapid development of laser scanners has created new areas of research. Laser scanning, through the non-contact scanning of surfaces, provides the basis for the reconstruction of objects whose shape is unavailable, lost or denied. Scanning devices mounted on robot arms enable the automation of point cloud acquisition for the subsequent shape determination and reconstruction of objects.This thesis presents an automated scanning approach for dense point cloud acquisitions that allow the surface reconstruction and pose determination of objects using a robot arm and a tracked triangulation scanner. The objective of this work is twofold. Primarily, a method was developed that automatically detects objects, generates robot trajectories for a dense point cloud acquisition, and reconstructs their surface. A grid-shaped proceeding of the scanning device is executed in a predefined space primarily in order to derive a coarse point cloud of the object. Based on this coarse point cloud, robot trajectories were generated that ensure scanning at optimal scan distance, which leads to a dense point cloud. The method was applied to five test objects, which were successfully reconstructed as 3D prints. The method provides reliable and promising results. Based on the results, the comparison of original and reconstructed objects reveals average deviations of less than 1 mm for this specific setup. A shrinkage of the 3D prints due to cooling down of the material is assumed, which leads to a size reduction of the reconstructed objects. On average, the 3D prints are smaller than the original objects of up to 2%. A scale deviation is supposed due to the shrinkage of the 3D print. The second aim of this work was the automated pose determination of wooden workpieces in a higher-order coordinate system based on the developed method. The definition of a local object coordinate system enables the determination of transformation parameters in relation to a higher-order coordinate system. Robot paths were adjusted with respect to the object coordinate system for milling and drilling operations. According to a realised field campaign, the automated pose determination of the object achieves an accuracy of less than 2 mm in relation to the higher-order coordinate system. This accuracy is considered sufficient with regard to the execution of milling and drilling operations.