Real-Time Anomalous Activity Detection in Surveillance Videos

With the exponential growth of surveillance video data, there is a need for intelligent and efficient real time anomalous activity detection systems. In this paper, a robust spatio temporal auto encoder framework that takes advantage of spatial structure as well as temporal dynamics in video sequenc...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 7th International Conference on Intelligent Sustainable Systems (ICISS) S. 637 - 642
Hauptverfasser: Susitra, D, Reddy, Sathi Abhinay, Prasanth, Sajjala, Dhanalakshmi, K, J, Sylvia Grace, Shamreen Ahamed, B
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 12.03.2025
Schlagworte:
Accuracy
Anomaly detection
computer vision
Critical infrastructure
dynamic thresholding
real-time analysis
Real-time systems
Spatio-temporal autoencoders
Streaming media
Streamlit interface
Surveillance
surveillance videos
Thermal sensors
Thresholding (Imaging)
Video sequences
Visual analytics
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:With the exponential growth of surveillance video data, there is a need for intelligent and efficient real time anomalous activity detection systems. In this paper, a robust spatio temporal auto encoder framework that takes advantage of spatial structure as well as temporal dynamics in video sequences and dynamic thresholding for anomaly detection is developed. The system analyzes reconstruction errors to separate normal from anomaly activities. Dynamic thresholding is incorporated to increase adaptability in varying environmental conditions and to decrease the number of false positives typical in static systems. The proposed solution is demonstrated with a Streamlit based user interface, which provides an interactive graph to show real time results, tracking anomalies as well as its customizable parameter configuration. Experimental evaluations are presented on benchmark datasets and shown better detection accuracy and faster response time than the conventional methods. A fundamental leap in automated surveillance, this system empowers secure protection of varied environments ranging from urban monitoring, public spaces, and critical infrastructure.
DOI:10.1109/ICISS63372.2025.11076326