A new algorithm to detect small indentations up to 10 gf, optimizing image processing time to determine Vickers hardness

Abstract We present a novel image processing algorithm to detect vertices in Vickers indentations, with areas of the order of tens of square micrometers, generated by forces of 10 and 50 gf. The detection of these small indentations by image processing had not been previously published. The algorithm uses binarization segmentation, automatically adjusting the threshold using standard histogram equalization for each input image. We used geometric parameters of a rhombus and the Hough transform to identify traces of indentations even smaller than the objects generated by the noise in the image. Our algorithm determines the four vertices of the indentation from the extreme points provided by the MATLAB regionprops function. We applied the algorithm to 60 very small indentations in high-noise images, corresponding to samples of titanium nitride, hafnium nitride, titanium-coated 316 stainless steel, and titanium nitride and hafnium nitride bilayers, obtained by a microhardness machine. The proposed algorithm applied to the 60 images presents a relative error of less than 1% compared to the results obtained by a conventional manual procedure. In addition, when applying the algorithm to 230 previously reported indentation images, a similar relative error was obtained, but with a significantly shorter processing time.