Intelligent Video Surveillance Systems with Violence Detection

Abstract

Smart video surveillance systems are increasingly becoming mandatory in the provision of security in the social arenas like schools, transport stations, shopping centers and in the streets. Old-fashioned surveillance systems are highly dependent on manual surveillance that may not be very efficient and easily subject to human error. In order to overcome this problem, this project will introduce a proposal of Intelligent Video Surveillance System with Violence Detection based on the latest deep learning methods. The system is based on the YOLOv8 object detection model that supports the analysis of video streams automatically and the identification of violent activities in real time. The system is able to identify suspicious behavior like fighting or violent behavior and categorize them as possible violence events through processing of the frames that the surveillance cameras have captured. The designed system will follow a systematic pipeline which integrates the dataset collection process, preprocessing, annotation, model training, and real-time detection. Whenever violentness is detected, a system will initiate auto-alarm system like powering on of the buzzer or email notification to authorities. This will enable responding faster and being able to have a better situational awareness. The experimental findings indicate that the system is capable of identifying violence very accurately in real-time and is therefore appropriate to be applied in smart surveillance systems. The solution helps in improving the safety of the people by increasing the scope of automated surveillance and quick response in an emergency.

Introduction

The text discusses the need for advanced surveillance systems due to rising violence and the limitations of traditional video monitoring, which relies heavily on human observation and is prone to errors and inefficiency. To address this, the proposed system uses artificial intelligence and computer vision to automatically detect violent behavior in real time.

The system is based on the YOLOv8 deep learning model, which processes video frames from surveillance cameras, performs preprocessing, and identifies violent activities such as fights or aggressive movements. When such behavior is detected, the system generates immediate alerts (e.g., alarms or notifications), reducing reliance on manual monitoring and enabling faster response.

The methodology includes collecting and preprocessing video datasets (both violent and non-violent), training the model, and designing a multi-layer architecture consisting of input, processing, decision-making, alert, and output layers. The system emphasizes real-time detection, accuracy, and scalability.

Experimental results show that the system achieves high performance, with around 92% detection accuracy, fast response time, and better precision and recall compared to basic models. It also improves user experience through an intuitive interface and reduces the workload of security personnel.

Conclusion

The Intelligent Video Surveillance System with Violence Detection is an efficient way to enhance security monitoring in a public setting. Conventional surveillance systems are very manual and subject to human mistakes and ineffectiveness. The offered system unites the deep learning and computer vision algorithms allowing it to analyze video streams and identify violent activities in real time. The system can detect suspicious behavior like fights or aggressive behavior with a high precision and rate with the usage of the latest models like YOLOv8. According to the results of the experiment, the suggested system works effectively in terms of detecting violent situations, and at the same time, it does not lose the ability to process in real time. The system is able to analyze video frames, detect possible

References

[1] V. D. Huszár, V. K. Adhikarla, I. Négyesi, and C. Krasznay, “Toward Fast and Accurate Violence Detection for Automated Video Surveillance Applications,” IEEE Access, vol. 11, pp. 18772–18793, 2023. [2] P. Negre, M. Noguera, and J. L. Martinez, “Deep Learning-Based Detection of Physical Violence in Surveillance Videos: A Systematic Review,” IEEE Access, vol. 12, pp. 43215–43234, 2024. [3] M. A. Etemadi Naeen, H. Mohammadzade, and S. B. Shouraki, “Human-Centric Anomaly Detection in Surveillance Videos Using YOLO-World and Spatio-Temporal Deep Learning,” IEEE Access, vol. 13, 2025. [4] P. Nardelli and D. Comminiello, “JOSENet: A Joint Stream Embedding Network for Violence Detection in Surveillance Videos,” IEEE Access, vol. 12, pp. 72010–72024, 2024. [5] S. A. Jebur, K. A. Hussein, H. K. Hoomod, L. Alzubaidi, and Y. Gu, “A Scalable Deep Learning Framework for Anomaly Detection in Surveillance Videos,” IEEE Access, vol. 12, pp. 90412–90428, 2024. [6] T. Aremu, Z. Li, R. Alameeri, M. Khan, and A. El Saddik, “SSIVD- Net: Salient-Super Image Classification for Weaponized Violence Detection in CCTV Videos,” IEEE Access, vol. 10, pp. 84561–84574, 2022. [7] P. M. Sethi and S. R. Babu, “Deep Learning-Based Violence Detection Using YOLO-V7 for Smart Surveillance Systems,” in Proc. IEEE Int. Conf. Smart Computing, 2025. [8] Baala, M. Hanoune, and M. Mohssine, “Deep Learning Approaches for CCTV-Based Violence Detection: A Survey,” IEEE Access, vol. 13, pp. 101233–101251, 2025. [9] M. R. Islam, A. B. M. Zaman, and J. Majumder, “Crime Prediction by Detecting Violent Objects and Activities Using YOLOv8 and MoViNet,” in Proc. IEEE Int. Conf. Artificial Intelligence Systems, 2024. [10] V. Janani and S. Rajkumar, “Deep Learning-Based Human Violence Detection in Surveillance Videos Using Spatio-Temporal Networks,” IEEE Access, vol. 13, 2025. [11] S. Sung and J. Park, “Design of Intelligent Video Surveillance System for Crime Prevention Using Deep Learning,” IEEE Access, vol. 10, pp. 88216–88228, 2022. [12] K. Merit, M. Khan, and A. Hussain, “AI-Based Violent Incident Detection in Surveillance Videos Using ResNet and BiGRU Networks,” IEEE Access, vol. 13, pp. 23041–23056, 2025. [13] S. Battiato, F. Puglisi, and G. M. Farinella, “Deep Learning Methods for Real-Time Crime Detection in Surveillance Video Streams,” IEEE Access, vol. 11, pp. 55890–55905, 2023. [14] M. Liu and Z. Xue, “Evaluation of Security Video Surveillance Systems Using Fuzzy Multi-Level Analysis,” IEEE Access, vol. 10, pp. 118540–118552, 2022. [15] S. Kumar, A. Patel, and R. Singh, “Real-Time Violence Detection in Surveillance Videos Using CNN-LSTM Architecture,” IEEE Access, vol. 12, pp. 69812–69826, 2024.

Copyright

Copyright © 2026 Mekala Goutham Reddy, Bella Pavan Kumar, B. S. Kiruthika Devi . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.