DEVELOPMENT OF A DYNAMIC PARTITIONING MODEL AND A SOFTWARE SYSTEM FOR IMAGE SEGMENTATION IN THE IMPLEMENTATION OF COMPUTER VISION FOR ROBOTIC SYSTEMS

Abstract

The article presents an approach to image segmentation using modern mathematical and software tools. In particular, it applies the apparatus of the theory of optimal set partitioning in dynamic formulations, as well as modern languages, technologies, and software development tools for implementing image segmentation algorithms and methods. The work emphasizes the use of the obtained results in training robotic systems, which can be employed for mitigating the consequences of man-made disasters.

The methods and algorithms of the theory of optimal set partitioning are used to formalize the problem, identify key factors and segmentation objects, and specify the components of the objective functional. The software implementation of the methods and algorithms is carried out using modern programming languages, technologies, databases, and database management systems.

The article provides a sufficient number of computer experiment results, which clearly demonstrate the adequacy of the applied models, methods, and their algorithmic implementations. The scientific research outcomes have been tested on real land-based unmanned devices. The article outlines both the identified advantages and shortcomings.

The main results of the study include the application of the proposed models for image segmentation aimed at isolating individual objects in photos and videos. The paper systematizes and classifies approaches to image segmentation. A comparative analysis of classical and AI-based methods is performed, based on experimental applications to images simulating emergency conditions.

The study substantiates the appropriateness of using U-Net in cases where annotated data are available, and Watershed for resource-constrained devices without the need for training. The results enable flexible selection of tools tailored to the specific nature of the task.

The aim of the research is to develop a mathematical model and methods for implementing graphical image segmentation, as well as a software system for executing the mentioned algorithms and obtaining segmented images.

The research was conducted on a personal computer with the following configuration: Intel Core i7-12700K, 8 cores, 3.6 GHz; RAM: 32 GB, DDR4 3200 MHz; HDD: 2 TB.

The obtained results demonstrate the high accuracy of the proposed mathematical model and the correct implementation of the algorithms, as confirmed by the graphical output. The root mean square relative error did not exceed 6%.

The model is not only theoretically grounded but also practically suitable for integration into digital monitoring and control systems for autonomous robotic platforms. The mathematical model and software application have been practically tested and have shown high efficiency and accuracy.

Keywords: mathematical model, robotic systems, image segmentation, optimal set partitioning theory, clustering, classification.