Smart Surveillance System Leveraging Machine Learning Techniques
Keywords:
Smart Surveillance, Video Classification, InceptionNet, Gated Recurrent Units (GRU), Activity Detection, Deep Learning, Real-time Monitoring, Threat Detection, Violence DetectionAbstract
This project aims to develop an intelligent surveillance system capable of real-time video classification and activity detection to enhance public safety and threat prevention. The system categorizes video streams into three main classes: Normal, Violence, and Weaponized scenarios. To achieve this, a hybrid deep learning model is employed, integrating InceptionNet for spatial feature extraction and Gated Recurrent Units (GRU) for temporal sequence modeling. InceptionNet captures intricate spatial details from individual video frames, while GRU effectively learns the temporal patterns across sequences of frames. The system is trained on a comprehensive and diverse video dataset that simulates real-world conditions, ensuring robustness, high accuracy, and generalizability. Real-time monitoring and alert generation are core features of the proposed system, making it highly suitable for deployment in modern surveillance infrastructures. The project also investigates various optimization strategies to improve inference speed and efficiency, supporting practical implementation in real-time environments.
Downloads
References
T. Wang, A. Gupta, and J. Lin, "Real-time violence detection in surveillance videos using deep learning," IEEE Access, vol. 9, pp. 145203–145215, 2021, doi: 10.1109/ACCESS.2021.3120539.
S. Hossain, M. A. Rahman, and R. Islam, "Weapon detection using CNN and RNN-based hybrid deep learning," Journal of Visual Communication and Image Representation, vol. 80, pp. 103302, 2022, doi: 10.1016/j.jvcir.2021.103302.
N. Patel and R. Agarwal, "Hybrid models for human activity recognition in smart surveillance systems," International Journal of Computer Vision and Signal Processing, vol. 11, no. 2, pp. 45–53, 2022.
L. Zhang and Y. Chen, "Video classification using CNN-RNN for automated surveillance," Proceedings of the IEEE International Conference on Smart Computing (SMARTCOMP), 2020, pp. 180–185, doi: 10.1109/SMARTCOMP50058.2020.00038.
M. R. Ahmed, S. M. Reza, and H. A. Khan, "Violent action recognition in videos using temporal convolutional networks," Multimedia Tools and Applications, vol. 81, pp. 4213–4231, 2022, doi: 10.1007/s11042-021-11767-z.
Y. Li and C. Zhou, "Smart surveillance system based on deep learning for real-time activity detection," Sensors, vol. 22, no. 1, pp. 120–134, 2022, doi: 10.3390/s22010120.
A. N. Krizhevsky and G. E. Hinton, "ImageNet classification with deep convolutional neural networks," Communications of the ACM, vol. 60, no. 6, pp. 84–90, 2017, doi: 10.1145/3065386.
H. Tran and T. Nguyen, "Temporal modeling in human activity recognition using GRU networks," Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP), 2021, pp. 228–234.
L. Liu, W. Liu, and M. Z. Shih, "Real-time activity recognition in surveillance videos using CNN-LSTM fusion," IEEE Transactions on Circuits and Systems for Video Technology, vol. 31, no. 9, pp. 3485-3498, 2021, doi: 10.1109/TCSVT.2021.3058347.
J. Wang, D. Liu, and F. Zhang, "Human behavior recognition in surveillance videos using deep neural networks," Journal of Computer Vision and Image Understanding, vol. 198, pp. 102973, 2021, doi: 10.1016/j.cviu.2020.102973.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 International Journal of Scientific Research in Computer Science, Engineering and Information Technology
This work is licensed under a Creative Commons Attribution 4.0 International License.