Detection of noncircularity and eccentricity of a rolling winder by artificial vision

A common objective in the web transport industry is to increase the velocity as much as possible. Some disturbances drastically limit this velocity. Time-varying eccentricity of the rolling winder is one of the major disturbances which affect the quality of the rolling winder. This unsuitable factor can lead to a web break for a high-speed winding process. The main contribution of this work is to offer a new measurement technique that is able to provide on-line the estimation of the roll radius and its variations with a subpixel accuracy. A key feature within this work is the contour curvature classification by means of wavelets decomposition of the edge orientation function. We also propose a new model accounting for the increasing radius of the rolling winder, which confirms the experimental results and the reliability of the proposed approach.