DESIGNING AN IMAGE CLASSIFIER FOR FAST SEARCH IN LARGE-SCALE DATABASES

The rapid increase in the volume of information on the Internet necessitates the development of effective methods for its realtime processing in information systems. In particular, an important aspect is the clustering of news information, which considers the morphological analysis of texts and visual content. This creates a relevant challenge in clustering images accompanying textual information on various web resources, such as news portals, social networks, and informational websites. Image clustering allows for more efficient structuring of large data sets and simplifies searching for the necessary information. The subject of this study is the creation of an image classifier that is resistant to the rapid growth of information in databases. This is especially important in conditions where the amount of information is increasing at a high rate, and there is a need to quickly and efficiently process large data arrays. The study aims to enhance the performance of searching for identical images in large databases, where the rate of new information addition can reach 10–12 thousand images per day. This requires the development of a specialized image classifier that can provide fast and accurate clustering despite the intense growth in the volume of information. To achieve this goal, various modern methods are used, including mathematical modeling, contentbased image retrieval, image processing techniques, and decision-making methods. One of the main tools applied in the study is the two-dimensional discrete cosine transform, which allows for efficient image encoding and data compression without loss of important information. The research results demonstrate that the developed image classifier is indeed resistant to the increase in the amount of information in databases. The analysis of the classifier’s properties showed that the proposed solution ensures high data processing speed and minimal computational requirements. Experiments have proven that image clustering using this approach is relatively fast and resource-efficient. The developed classifier can significantly improve the efficiency of information systems, particularly in the context of the constant growth of information flows, making it an important tool for processing large image databases.