Segmentation and Fitting for Geometric Reverse Engineering : Processing data captured by a laser profile scanner mounted on an industrial robot

Abstract: Geometric Reverse Engineering (GRE) is the problem of creating CAD models of real objects by measuring point data from their surfaces. The GRE process is usually divided into four sub processes, data capturing, preprocessing, segmentation, fitting and finally CAD model creation. To automate the GRE process the sub processes need to be integrated. The work presented in this thesis concerns the integration of the sub processes and shows how to implement GRE using a system consisting of a laser profile scanner mounted on an industrial robot equipped with a turntable.The first paper is concerned with the relatively complex accuracy issues of the system and the sources of errors of the system are identified.The second paper continues the work with accuracy issues to investigate the 2D accuracy of the scanner head. Moreover it presents an implementation of planar segmentation and fitting based on point clouds. A new method for planar segmentation based on a laser profiles and robot poses is proposed.The Third paper presents a comprehensive  review of the recent development of the measurement system and discusses the integrated GRE process and the requirements for integrating the GRE sub processes using an open source CAD tool which is maintained and developed at Örebro University.The fourth paper addresses the problem of segmentation of the measured data into planar regions by implementing and comparing two different algorithms. The first algorithm is well known and based on point clouds. The second algorithm is new and based on laser profiles and robot poses. Experimental results indicate that the second algorithm is faster and more accurate.The final paper continues the work on segmentation and fitting. A new algorithm to solve the complex problem of quadric segmentation and fitting is presented.Finally, the thesis points out interesting directions of future work.