Proceedings of the Polymer Processing Society 26th Annual Meeting ~ PPS-26 ~ July 4-8, 2010 Banff (Canada) NOVEL METHODOLOGY FOR THE IN-LINE MEASUREMENT OF THE VISUAL PERCEPTIBILITY OF SURFACE DEFECTS - 100 PERCENT QUALITY CONTROL D.P. Gruber 1 , J. Macher 1 , D. Haber 1 , W. Friesenbichler 2 1 Polymer Competence Center Leoben (PCCL), Roseggerstraße 12, 8700 Leoben, Austria – dieter.gruber@pccl.at 2 University of Leoben, Franz Josef Straße 18, 8700 Leoben, Austria – walter.friesenbichler@unileoben.ac.at Uniformity of surface gloss and surface structure of products are important industrial demands. Thereby it is to consider that human vision is very sensitive when it comes to the recognition of surface defects like weld lines, sink marks or areas of inhomogeneous gloss. While there is broad agreement on obvious surface defects, evaluation becomes complex if defects are close to the threshold of visibility. For that reason there is a strong demand of objective and reproducible measurement of the visual perceptibility of surface defects. A novel methodology for the advanced in-line measurement of surface defect visibility was developed at the PCCL. The methodology addresses the given issue by including knowledge about human perception behavior in the mathematical measurement models. The basis of a product’s appearance is light reflected from the surface to the eye-apparatus where humans get the impression of color and contrast shades. Depending on contrast perception characteristics surface defects disturb the contrast impression of uniform surfaces and surface structures. Human perception like characterization of surface defects has been achieved by implementing contrast perception characteristics in the form of mathematical models into measuring methods. Measurement values are calcualted by the application of the model algorithms to images captured by an industrial CCD- camera. By this means results are unambiguous and reproducible while there is no subjective influence of human evaluation. In this paper the characterization of a sink is presented. An inspection unit was developed for on-line inspection of injection molding parts. Test samples were prepared in cooperation with the University of Leoben. Thereby, the process parameters were varied during the fabrication of series of samples. The measurement setup was adjusted to acquire images of the sink mark with good contrast in order maximize the accuracy of the characterization process. The intensity matrices of the images were processed using a mathematical model which allows prediction of the visibility of the defect. The results of machine vision showed good correlation with the results of human assessment. Introduction Besides performance and handling, appearance has become one of the main features of an increasing number of products e. g.: mobile telephones or LCD- monitors. Therefore, there is an industrial requirement to control and, if necessary, enhance consumer products during the manufacturing processes [Pointer, 2003]. Especially the control of visual appearance of product surfaces is still conducted by human evaluators. The work is time consuming as well as subjective and dependent on the evaluators´ experience and daily condition. Machine vision applications for defect detection overcome these problems by ensuring continuous and objective on-line inspection at lower costs. [Guzaitis, 2006], [Yan-Hsin Tseng, Du- MingTsai, 2010]. At the Polymer Competence Center Leoben GmbH (PCCL) methodologies were developed which reproduce the perception of surfaces by the human eye apparatus [Gruber, 2006]. Algorithms were implemented into an inspection system for on-line inspection of injection molding parts and are presented in this paper by using the example of a sink mark as a frequent defect. High defect reliability and low computational cost were the sought qualities to ensure the practicability of the inspection system for on-line inspection. Samples for the tests of the inspection system were manufactured in cooperation with the University of Leoben, while it was of particular interest how the testing site would respond to the subtle differences of the surface defects of samples which were processed with different processing parameters. Therefore, the parameters were varied during the manufacturing to obtain an overview of their influence on the characteristics of the surface defects. Experimental Setup The test samples were produced using injection molding. Process parameters like melt and mold temperature as well as injection velocity and holding pressure were varied to obtain an overview of their influence on the characteristics of the surface and its defects. The limits of these variations were chosen according to a processing window where the mold was totally filled without any leakage out of the mold. A well defined defect which disappears within the chosen parameter window was chosen. Figure 1 shows the section of the injection molding part with the defect which was analyzed in the first tests. The defect (square), a punctual sink mark, also interacts with a