Methodology for building a comprehensive logical model for detecting potentially dangerous situations using computer vision technologies

The relevance of the study stems from the need to overcome the fragmentation of existing video-based hazard detection systems, which are designed as highly specialized solutions for specific scenarios, thereby hindering their scalability and adaptation to evolving safety requirements. Accordingly, the aim of this work is to develop a unified methodological approach to building a universal computer vision system for automated threat recognition. The leading research method was conceptual modeling: by abstracting and grouping typical violations (ranging from traffic rules to industrial regulations), a limited set of basic visual concepts ("Human", "Danger Perimeter", "Mechanism", "Signaling Device", "Personal Protective Equipment") and their states was identified. To formalize the relationships between these concepts, a relation matrix was introduced, and the logic of hazard detection is described through formal logical formulas covering two analysis zones: a static zone that operates with instantaneous Boolean features, and a dynamic zone that applies temporal logic to predict the evolution of events. As a result, a conceptual model has been created that incorporates a universal formal apparatus capable of describing a wide spectrum of threats as a system of logical rules adjustable to specific scenarios. Given the urgency of digitalization in industrial safety, the materials of the paper are of practical value for software developers in the field of occupational safety. The approach proposed in this work enables scalable and flexible tuning of monitoring systems for specific facilities.

1. Marenkov A.N., Kuznetsova V.Yu., Gelagaev T.M. Application of face recognition technologies in control and access control systems. Caspian Journal: Management and High Technologies. 2021;(1):83–90. (In Russ.). https://doi.org/10.21672/2074-1707.2021.53.1.083-090

2. Morozov A.A., Sushkova O.S. Real-time analysis of video by means of the Actor Prolog language. Computer Optics. 2016;40(6):947–957. (In Russ.). https://doi.org/10.18287/2412-6179-2016-40-6-947-957

3. Artikis A., Sergot M.J., Paliouras G. A Logic Programming Approach to Activity Recognition. In: Proceedings of the 2nd ACM International Workshop on Events in Multimedia (EIMM 2010), 25–29 October 2010, Firenze, Italy. New York: ACM; 2010. P. 3–8. https://doi.org/10.1145/1877937.1877941

4. Goryaev V.M., Mankaeva S.A., Sumyanova E.V., et al. Optimizing Hybrid Long-Term Recurrent Convolutional Network Architecture for Edge-Based Violence Detection in Video Surveillance Systems. Modern Science and Innovations. 2025;(3):30–38. (In Russ.). https://doi.org/10.37493/2307-910X.2025.3.3

5. Leus A., Tatarinova E., Efremov V., et al. A system for multi-camera temporal identification and tracking of a person at the enterprise. Legal Informatics. 2022;(3):31–39. (In Russ.). https://doi.org/10.21681/1994-1404-2022-3-31-39

6. Sergeev N.E., Tselykh Ju.A. Using temporal relations to describe complex scenes based on video images. Izvestiya SFedU. Engineering Sciences. 2009;(12):134–142. (In Russ.).

7. Barbosa F.G.O., Mourao G.L., Erazo J.J.M., et al. Dynamic risk management in office workspaces: Real-time analysis of staircase safety using computer vision and AI. Computers and Electrical Engineering. 2025;122(2):109902. https://doi.org/10.1016/j.compeleceng.2024.109902

8. Kashevnik A., Othman W., Ryabchikov I., et al. Estimation of motion and respiratory characteristics during the meditation practice based on video analysis. Sensors. 2021;21(11):3771. https://doi.org/10.3390/s21113771

9. Lee J., Lee S. Construction site safety management: A computer vision and deep learning approach. Sensors. 2023;23(2):944. https://doi.org/10.3390/s23020944

10. Chumakov K.V., Strizhak P.A., Kropotova S.S. Using video analytics to monitor and improve the safety of workflows. Occupational Safety in Industry. 2025;(1):83–89. (In Russ.). https://doi.org/10.24000/0409-2961-2025-1-83-89

11. Orlov A.A., Ryzhkova M.N., Mareev A.V. Methods of object localization in video stream. Radioengineering and Telecommunication Systems. 2021;(3):48–60. (In Russ.).

12. Bouwmans Th. Traditional and recent approaches in background modeling for foreground detection: An overview. Computer Science Review. 2014;11–12:31–66. https://doi.org/10.1016/j.cosrev.2014.04.001

13. Dosalieva A.F., Demina R.Y. Comparative analysis of synthetic video generation platforms: Finding the optimal solution for security issues. In: Artificial intelligence: Scientific achievements and applied tasks: Materials of the 2nd All-Russian scientific and practical conference with international participation, 04–05 December 2025, Khanty-Mansiysk, Russia. Khanty-Mansiysk: Yugra State University; 2026. P. 26–32. (In Russ.).