Detekce anomálií v reálném čase a odhad hustoty davu v uzavřených prostorách pomocí umělé inteligence

• Thesis title in Czech: Detekce anomálií v reálném čase a odhad hustoty davu v uzavřených prostorách pomocí umělé inteligence
• Thesis title in English: Real-time anomaly detection and crowd density estimation in enclosed spaces using AI
• English key words: real-time person detection|crowd density estimation|deep learning
• Academic year of topic announcement: 2025/2026
• Thesis type: diploma thesis
• Department: Department of Software and Computer Science Education (32-KSVI)
• Supervisor: doc. RNDr. Tomáš Dvořák, CSc.

Guidelines

Important note: This topic will be supervised by Daniel Mai, Assistant Professor in Computer Science, Efrei Research Lab, Paris Panthéon-Assas University, within the 4EU+ Alliance.

The goal of this thesis is to expand on the real-time person detection and counting system by integrating crowd density estimation and anomaly detection for enclosed spaces such as laboratories or restricted areas. This enhanced system will utilize video feeds from surveillance cameras to not only monitor occupancy but also identify unusual crowding patterns that may indicate potential safety concerns or over-occupancy. By leveraging deep learning algorithms in real-time, the system will analyze the movement patterns of individuals, flagging sudden changes in density or abnormal behaviors (e.g., rapid influx of people in restricted zones) that may necessitate immediate intervention.

This project requires developing a model using Python and deep learning frameworks such as OpenCV and CNNs. The model will involve training and customizing object detection and tracking algorithms to account for different crowd levels and movement patterns, adapting to indoor environments where movement is often dynamic. In addition to real-time crowd estimation, the system should be capable of alerting security systems when occupancy levels approach or exceed safe limits. Performance testing will be conducted to ensure the model's accuracy and responsiveness in various scenarios.

References

Rezaee, K., Rezakhani, S. M., Khosravi, M. R., & Moghimi, M. K. (2024). A survey on deep learning-based real-time crowd anomaly detection for secure distributed video surveillance. Personal and Ubiquitous Computing, 28(1), 135-151.
Bamaqa, A., Sedky, M., Bosakowski, T., Bastaki, B. B., & Alshammari, N. O. (2022). SIMCD: SIMulated crowd data for anomaly detection and prediction. Expert Systems with Applications, 203, 117475.
Gao, G., Gao, J., Liu, Q., Wang, Q., & Wang, Y. (2020). Cnn-based density estimation and crowd counting: A survey. arXiv preprint arXiv:2003.12783.
Sindagi, V. A., & Patel, V. M. (2018). A survey of recent advances in cnn-based single image crowd counting and density estimation. Pattern Recognition Letters, 107, 3-16.

Preliminary scope of work in English

The goal of this thesis is to expand on the real-time person detection and counting system by integrating crowd density estimation and anomaly detection for enclosed spaces such as laboratories or restricted areas.
This topic will be supervised by Daniel Mai, Assistant Professor in Computer Science, Efrei Research Lab, Paris Panthéon-Assas University, within the 4EU+ Alliance.