Efficient borehole detection from single scan data |
| |
Authors: | Markus Vincze Walter Wohlkinger Georg Biegelbauer |
| |
Affiliation: | (1) School of Informatics, University of Edinburgh, James Clerk Maxwell Building, Mayfield Road, EH9 3JZ Edinburgh, UK;(2) Artificial Intelligence Center, SRI International, 333 Ravenswood Ave., 94025 Menlo Park, CA, USA |
| |
Abstract: | This work presents a fast and robust method to precisely segment and locate boreholes of 4–50 mm diameter. The task is solved
by scanning over the expected borehole location given by CAD data. Since the diameter of the borehole is known, this can be
used to obtain a robust and fast algorithm suitable for industrial application. Single scan data is sufficient to segment
the bore independent of bore chamfer type using a robust normal vector fit and a classification based on the Gaussian image.
A sequential cylinder fit algorithm makes it possible to calculate the bore pose in less than one second. The experiments
demonstrate that 120° of the borehole surface are sufficient for robust localization within 0.3 mm and 0.5° even in the presence
of ghost points and notches in the boreholes. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |
|