ТРИРІВНЕВА СИСТЕМА ВЕРИФІКАЦІЇ ГРАФІЧНОГО КОНТЕНТУ НАУКОВИХ ПУБЛІКАЦІЙ НА ОСНОВІ ТОПОЛОГІЧНОГО АНАЛІЗУ ДАНИХ

Object of research is the process of verifying graphical content in scientific publications. Subject of research is methods and algorithms for detecting plagiarism of graphical information based on data content analysis. The aim of the work is to develop and substantiate a three-level system for verifying graphical content in scientific publications. Research results. The paper proposes a novel comprehensive method for detecting plagiarism of graphical information in scientific publications. Unlike existing systems (Proofig, Imagetwin) that detect only visual duplicates, the proposed approach analyzes the content of graph data. The method is based on a combination of three levels of analysis: reverse engineering of numerical data from graphical images, geometric analysis of curve shapes using the Dynamic Time Warping (DTW) algorithm, and Topological Data Analysis (TDA) using persistent homology and Wasserstein distance. The scientific novelty lies in the application of topological invariants for comparing the content of graphical data, which ensures robustness against visual manipulations: changes in axis scale, color scheme, line style, and label language. An integrated metric for assessing graph similarity has been developed. The effectiveness of the approach is confirmed by an example of detecting masked plagiarism.