Архітектура мультимедійної системи персоналізованої навігації на основі семантичної розмітки об’єктів

Abstract

This paper proposes a novel framework for personalized navigation in complex multimedia environments. By integrating Computer Vision (YOLO) and OCR (CRNN), the proposed approach constructs a dynamic knowledge graph of exhibits, moving beyond static topological maps. Path costs are defined through a multiobjective linear scalarization function, balancing semantic affinity with spatial distance. This transition from geometric pathfinding to context-aware, interest-based exploration significantly enhances the information density of user visits. The method demonstrates high efficiency in real-time path optimization, offering a scalable solution for intelligent guidance systems in visually dense spaces.