Enhancing Surveillance Systems leveraging AIoT for Advanced Object Detection in Real-Time Security Applications

Surveillance systems are integral for modern security frameworks, ensuring safety in public areas, traffic monitoring, and infrastructure protection. The rising demand for surveillance necessitates advanced technologies capable of accurately identifying objects such as vehicles and pedestrians in real time. A promising solution is the use of AI technologies that has proven their ability in object detection with exceptional accuracy. This study proposes an optimized object detection model using YOLOv8, incorporating comprehensive hyperparameter tuning through the Optuna framework, significantly enhancing object detection accuracy and response time in dynamic environments. The model was trained on the PASCAL VOC 2012 dataset. Our results highlight a robust model performance, achieving a precision of 86.2%, a recall of 78.9%, a mean Average Precision (mAP) of 84.7%, and an F1-score of 81%, which surpasses previous benchmarks for this dataset, particularly in challenging crowded scenarios. Integrating the Internet of Things (IoT) with AI forms Artificial Intelligence of Things (AIoT), a revolutionary step that leverages interconnected devices for real-time, automated data processing and rapid response in complex scenarios, addressing scalability and responsiveness in security applications. Our model detection results are transmitted to the ThingSpeak IoT platform, enabling efficient real-time monitoring and analysis, demonstrates the feasibility of using AIoT to develop scalable, high-accuracy surveillance systems with potential applications across urban safety, traffic management, and infrastructure security.