1077-2618/13/$31.00©2013IEEE S STEEL STRIPS PRODUCED IN steel works are used as raw material in many other industries, control of their quality is essential. Steel is required for the production of many products such as tools, cans, or car parts. In all of these cases, the quality of the steel has a direct impact on the quality of the final product. One of the most critical phases of steel quality control takes place in the finishing mill, where the hot steel is rolled into its final form. If the roll surface that applies pressure on the steel has any kind of distortion, it imprints a set of defects on the steel strip. Each time it completes a turn, it gener- ates a defect. This is a crucial problem. This article proposes a technique to detect the defects caused by defective work roll. It includes a way to store the information necessary for detection as well as a flexible algorithm that uses this information efficiently. To train the algorithm to obtain the best possible outcome for a set of test strips, a way to quantify each solution is also proposed. Finally, the results obtained are compared with those obtained by a commercial tool, and the improve- ment achieved using the new technique is discussed. Detection through a vision-based technique By FRANCISCO G. BULNES, RUBÉN USAMENTIAGA, DANIEL F. GARCÍA, JULIO MOLLEDA, & JOSÉ L. RENDUELES Digital Object Identifier 10.1109/MIAS.2012.2215638 Date of publication: 26 December 2012 © DIGITAL VISION A 39 IEEE Industry ApplIcAtIons MAgAzInE • MAr|Apr 2013 • www.IEEE.org/IAs