Resumen
The article shows that the development of an effective segmentation algorithm for metallographic images is an urgent task. The mean shift algorithm and its disadvantages are considered. To eliminate the shortcomings, a modification of the algorithm based on agglomerative clustering is proposed. This algorithm is as follows: at the first stage, standard clustering is performed by the mean shift method with a small window size; then the power of each point that is the center of the cluster is calculated, and a new set is formed from these points; then clustering is performed using the standard mean shift method of the resulting set. The last three stages are repeated until the clusters stop merging. An algorithm for finding deformation zones is also proposed. In order to reduce the time spent on their search, it is assumed to compare the contours of individual grains of metal or alloy with a circle or ellipse. As part of the implementation of the algorithms described in the article, it is proposed to create a software product consisting of two components: a component for searching grain boundaries and a component for detecting deformation and deformation zones. The software architecture is described for each component, and the test results are presented. For the software components presented in the article, 2 author's certificates of computer program?s registration were obtained.