IoT-Aided Smart City Architecture For Anomaly Detection
Jiaojie Yuan, Jiewen Zhao · International Journal of Critical Infrastructures · 2024
Anomaly detection in smart cities is critical for mitigating human fall-related injuries and fatalities, particularly within IoT devices. Despite numerous vision-based fall detection methods, challenges persist regarding accuracy and computation costs, especially in resource-constrained IoT environments. This paper proposes a novel fall detection approach leveraging the Yolo algorithm, known for its efficiency in minimising computation costs while maintaining high accuracy. By utilising a diverse fall image dataset, the method undergoes rigorous training and evaluation, employing standard performance metrics. The results reveal impressive precision, recall, and mean average precision (mAP) values of 93%, 89%, and 95%, respectively. Notably, the Yolo algorithm's computational efficiency ensures minimal resource utilisation, making it suitable for real-time deployment in IoT devices within smart city infrastructures. Consequently, this method presents a promising solution for enhancing fall detection accuracy while optimising computational resources, thus advancing safety measures in urban environments.