PSI - Issue 72

Levente Tatár et al. / Procedia Structural Integrity 72 (2025) 345–353

348

2.1. Optical measurement method For optical measurements an in-house built system has been used. The system used a JAI GO-5000-USB camera with 2560 x 2048 pixels resolution. The method of processing image data is presented schematically in Fig. 2.

Fig. 2. Schematics of processing raw graphical data

In this article we use only data obtained from specimen contours and we disregard surface markings. Operations on pictures have been done with the help of ImageMagick (2025) and in-house programs. Instead of using one single monolithic program, we used a succession of programs to do just one step at once. This provided more control over the result. First a rough cutting has been applied to the original picture to exclude parts which are irrelevant for image processing. Then the image was converted from greyscale (256 shades of gray) to black and white (1 bit palette). See Fig. 2a. The threshold for this conversion is crucial. A histogram has been used for the black and white picture along the axis of the specimen, summing up the number of black pixels for each line perpendicular to the axis. This way the extensometers could be easily detected, and further cutting has been applied (Fig. 2b). Then the ImageMagick “morphology” command has been used to create a black and white contour for the specimen. Vectorization of the bitmap picture has been done with the aid of our own program. Its philosophy is different from generally available vectorization programs. It divided the picture into two halves, so only one contour was contained in one half. It converted black pixels along each scan line perpendicular to the axis to a floating point number. As a final result of this step we obtained a 3 column data: scan line number, average coordinate (floating point) of pixels for one contour, average coordinate (floating point) of pixels for the other contour. The final step corresponding to Fig. 2d consisted of transformation of coordinates in pixels to real physical coordinates. For this step a calibration based on the initial dimensions of the specimen has been performed. 2.2. Measurement results Engineering strain – engineering stress and true strain – true stress curves had been created from data obtained from clip gauge and pictures. Filtering was necessary as raw data is noisy. Filtered data can be seen in Fig. 3.