Industrial applications like robot-aided welding, automated inspection, and 3D measurements require 3D points to be captured from the surfaces of objects and processed to calculate the information-of-interest. The lack of research focused on fitting ellipses to 3D laser profile data, and the intrinsic features that distinguish it from 2D digital images, motivated us to conduct a comparative study involving the most popular ellipse-fitting methods. After describing our laser profile scanning system, and a survey of ellipse-fitting methods, we compare, using extensive experiments performed with synthetic and real data, the fitting algorithms in terms of stability and accuracy with respect to a variety of factors. The estimate obtained with the best method is used to initialize a robust non-linear iterative ellipse fitting method. Finally, we describe a novel method for the construction of cylindrical surfaces from estimated elliptical sections.