Robust Circle Fitting in Industrial Vision for Process Control of Laser Welding
Nicolaj C. Stache, Henrik Zimmer · RWTH Publications (RWTH Aachen) · 2007
With the increase of computational power of standard personal computers, methods for robustly fitting analytically describable geometric shapes have become attractive for use in industrial image processing.One challenging task in this field of application is the monitoring of laser welding processes with the estimation of process parameters such as melt pool position and size.This has to be done with troughput rates beyond 200 frames per second in a noisy environment.For this purpose, we adapt two existing approaches for robust fitting to the dedicated task (RANSAC, Least Median of Squares estimation) and introduce a new technique, based on prior knowledge of the welding process.Extensive experiments yield that the new technique outperforms the existing approaches.