Data processing and annotation in a distributed video stream mining system to detect destructive behavior

The article discusses a distributed system for intelligent video stream analysis designed for automatic detection of destructive behavior in educational institutions. The relevance of the study is due to the need to improve the efficiency of security systems in organizations where existing systems demonstrate significant limitations in response speed and objectivity of assessment. The purpose of this work is to develop the architecture of a distributed system for intelligent video stream analysis based on a three-level video data processing pipeline to identify destructive behavior. The main focus is on methods for processing and annotating video data within a three-level pipeline, including object detection (YOLO), behavior classification (CNN) and contextual event analysis. The system based on the proposed architecture and modules of the three-level pipeline allows for effective detection of destructive behavior, which demonstrates the promise of using neural network technologies to create intelligent security systems in organizational structures. In this article, the authors consider the use of a distributed system for intelligent analysis of video streams in educational institutions, but the proposed solution can be adapted for other organizational structures where prompt detection of aggression, fights and other forms of destructive behavior in crowded conditions is required.