Development of Efficient Algorithms for Detecting Graphic Primitives on 2D/3D Images

This graduate qualification work deals with the development of algorithms for detecting graphic primitives in 2D/3D images. The research was carried out within the framework of cooperation between the National Research University Higher School of Economics and ViCueTech LLC for Huawei Tech Company LLC. The main goal of the work was to create an algorithm capable of detecting and analyzing contours in images, revealing their properties and possible defects. The introduction is devoted to describing the relevance of the problem, its industrial application and setting the research objectives. Chapter 1 contains a literature review, including different approaches to the problems of object detection, semantic segmentation and contour detection. Chapter 2 describes the developed operators to implement the proposed algorithm, including the basic signature and algorithm of the ApplyBeadInspectionModel operator. Chapter 3 focuses on testing the developed algorithm, analyzing the results and identifying errors, while Chapter 4 suggests potential improvements to improve the accuracy and efficiency of the algorithm. The conclusion summarizes the work done and its significance. The work is practical and demonstrates high results in contour detection, which is confirmed by experimental data and comparison with commercial analogs.