An Intelligent Spatiotemporal Transformer Framework for Secure Surveillance Video Analytics in Smart Cities

Analysis of surveillance video is critical in keeping the city safe in a smart city since it allows 24/7 camera control of the city environment. Such systems can be made smarter, more automated, and more responsive through the use of Artificial Intelligence. Most available methods struggle to balance effective anomaly detection with strong privacy assurance. Traditional models often rely on detailed visual information, which can compromise personal anonymity, yet are not reliable for identifying complex or subtle abnormalities in fluid crowd situations. These restrictions are lowering the system reliability, scalability, and public trust. A new model to overcome these issues; it is called Anomaly Detection in Public Spaces with Privacy-Aware Video Intelligence using Spatiotemporal Transformer Networks. The suggested system uses spatiotemporal transformers to capture both spatial and temporal dependencies in video streams. The model does not process raw, identifiable data; it uses selective feature extraction and privacy-preserving feature transformation methods. This ensures that detecting an anomaly does not require access to or storage of personally identifiable information. The system is anonymous yet effective at detecting behavioral patterns rather than identifying individuals. Suitable applications include crowd monitoring, abnormal event detection, and urban safety management. Experimental results show that the STN-based model significantly improves anomaly detection rates while ensuring greater privacy compared to current surveillance methods.