Digital image processing methods for the evaluation of optical anisotropy effects in tempered architectural glass using photoelastic measurements

In the present paper, optical anisotropy effects in architectural glass are evaluated using digital image processing. Hereby, thermally toughened glass panes were analyzed quantitatively using a circular polariscope. Glass subjected to externally applied stresses or residual stresses becomes birefringent. Polarized light on birefringent materials causes interference colors (iridescence), referred to as anisotropies, which affect the optical appearance of glass panes in building envelopes. Thermally toughened glass, such as toughened safety glass or heat strengthened glass, show these iridescences due to thermally induced residual stress differences. RGB-photoelastic full-field methods allow the quantitative measurement of anisotropies, since the occurring interference colors are related to the measured retardation values. By calibrating the circular polariscope, retardation images of thermally toughened glass panes are generated from non-directional isochromatic images using computer algorithms. The analysis of the retardation images and the evaluation of the anisotropy quality of the glass is of great interest in order to detect and sort out very low quality glass panes directly in the production process. Therefore, in this paper retardation images are acquired from different thermally toughened glass panes then different image processing methods are presented and applied. It is shown that a general definition of exclusion zones, e.g. near edges is required prior to the evaluation. In parallel, the limitations in the application of first-order statistical and threshold methods are presented. The intend of the investigation is the extension of the texture analysis based on the generation of Grey Level Co-occurrence Matrices, where the spatial arrangement of the retardation values is considered in the evaluation. For the first time, the results of textural features of different glass pane formats could be compared using reference areas and geometry factors. By reduction of the original image size, the computation time of textural analysis algorithms could be remarkably speeded up, while the textural features remained the same. Finally, the knowledge gained from these investigations is used to determine uniform texture features, which also includes the pattern of anisotropy effects in the evaluation of thermally toughened glass. Together with a global evaluation criterion this can now be implemented in commercial anisotropy measurement systems for quality control of tempered architectural glass.