EUROPEAN STEEL TECHNOLOGY AND APPLICATION DAYS 2017 - ESTAD 2017

**Abstract** In this work, we present a novel inspection technique for surface defects on cold and hot steel products during production. Common approaches for cold steel products are based on 2D images, which are highly influenced by the surface texture and thus lead to high false alarm rates on hot surfaces. Automated detection of defects on hot surfaces is of special interest, though, because it allows the manufacturer to react as early in production as possible and consequently minimize waste and labor costs. We propose the use of high-speed light section sensors that allow a complete 3D-Reconstruction of the inspected object, even when dealing with surfaces of up to 1000 degrees Celsius. Depth-Images are already delivered by the sensors and processed to so-called Residual-Images, that encode structural deviations on the surface. The Residual-Images are then segmented in order to detect and classify defects such as cracks and flakes. As this approach is based just on the 3D information given by the sensors, it is independent of the surface texture. It also allows calculating useful defect information such as depth and volume, as well as the calculation of common profile dimensions like product height and width. This makes it possible in the future to combine the surface inspection with high accuracy profile measurement into a multi-purpose system.