Go to JKU Homepage
Institute for Measurement Technology
What's that?

Institutes, schools, other departments, and programs create their own web content and menus.

To help you better navigate the site, see here where you are at the moment.

Texture Recognition at Coils

Patrick Hölzl


Quality assurance is essential in today's economy and will becoming a more important field increasingly. A large part of quality assurance processes use methods of digital image processing to optically evaluate the specimen. The problem which is worked on in this project also is in this category.

This project is part of a major project promoted by the Austrian Research Promotion Agency, opens an external URL in a new window and financed by the company Industrie-Logistics-Linz GmbH & Co KG (ILL), opens an external URL in a new window, with the purpose to improve the quality assurance for steel coils.

The main task of the project was to capture characteristic features on steel coils with the help of an optical measurement systems. The main objects visible on steel coils are the backup tapes, which can be horizontally and vertically arranged. Also packaging material and scratches that are caused by transport action or storage should be detected. The formats of the images that are used in this project are PNG and HDR. The processing is done in MATLAB, because of its extensive imageprocessing toolbox.

Figure 1: Steelcoil with detected objects; the coil position is marked blue, the tapes are marked green and the packaging material is marked red.

First of all the position of the coil must be detected within the frame. In order to detect the position of the coil in the image, a Canny edge detection algorithm in combination with pattern matching is used. The pattern matching process is based on mattched filters. Although the vertical coil position in the images is harder to estimate, due to the cylindrical coil form. Therefore the principle of central projection and the statistical properties of the coil form are used instead. As soon as the position of the coil is determined, the remaining object edges get a ssigned. The main problem for the object detection are the randomly occurring reflections in the image, because they generate additional object edges and thus will get detected falsely. Finally, using several programs written in MATLAB, detected objects on the coil will be marked (as shown in Figure 1).

April 20, 2009