One of the biggest aims of a logistics company is, apart from the optimization of processes, the internal quality assurance. Nowadays the quality management often gets support from modern technologies. Thereby digital image processing becomes increasingly important due to the availability of more powerful computation hardware and availability of better cameras. In this context the ultimate aim of digital image processing is to empower a machine to decide whether the product is of the desired quality or not based on one or more acquired images.
This work deals with one part of a project financed by the Österreichische Forschungsförderungsgesellschaft (FFG), opens an external URL in a new window in collaboration with the company Industrie-Logistik-Linz GmbH & Co KG (ILL), opens an external URL in a new window, which stores and ships steel coils manufactured by voestalpine AG, opens an external URL in a new window. The new idea is to unwrap the surface of the investigated steel coils virtually to provide a comparison with the condition of the coils at different instants in time and also to visualise damages such as dents, scratches on the surface and packing errors more easily. For this purpose mathematical models of the coils and of the camera are elaborated and modern methods for camera calibration and for estimating the model parameters are presented. After all algorithms for the virtual unwraping and possibilities of how to deal with problems like motion blur or non-homogeneous illumination of the scene in the individual partial images are going to be discussed. Finally the results are illustrated regarding processing of representative image series and subtraction images. An experimental acquisition setup has been made up at the Institute of Measurement Technology, where a pivoted replica of a coil allows simulated acquisitions for testing purposes.
Keywords: quality assurance, digital image processing, parameter estimation, geometrical rectifying of images.
The circulation of the project work is restricted from August 29, 2011 for a period of 5 years.
August 18, 2011