An IoT Enable Anomaly Detection System for Smart City Surveillance.

Area of Science:

• Computer Vision
• Artificial Intelligence of Things (AIoT)
• Surveillance Systems

Background:

• Smart cities generate vast surveillance data, necessitating automated anomaly detection.
• Traditional methods relying on human observation are tedious and inaccurate.
• Automated anomaly detection in complex, real-world surveillance is challenging.

Purpose of the Study:

• To present an efficient and robust framework for anomaly detection in large-scale surveillance video data using AIoT.
• To develop a hybrid model integrating 2D-CNN and ESN for smart surveillance applications.
• To enable lightweight, edge-deployable solutions for AIoT environments.

Main Methods:

• A hybrid model combining 2D-Convolutional Neural Network (CNN) for feature extraction and Echo State Network (ESN) for sequence learning.
• Utilized an autoencoder for feature refinement after initial CNN processing.
• Implemented the model on edge devices for real-time processing in AIoT settings.

Main Results:

• The proposed hybrid 2D-CNN and ESN model demonstrated significant performance enhancements.
• The framework proved effective in detecting anomalies within challenging surveillance datasets.
• The lightweight design ensures applicability and capability in AIoT environments.

Conclusions:

• The developed AIoT framework offers an efficient and robust solution for automated anomaly detection in smart city surveillance.
• The hybrid model successfully addresses the complexities of real-world anomaly detection in video data.
• Edge implementation makes the system practical for widespread AIoT adoption in smart cities.