Enhancing Real-Time Threat Detection with YOLOv8 and LSTM Integration for Intelligent Surveillance Systems

In the face of increasing security threats, efficient surveillance systems are crucial for timely threat detection and response. This study presents a comprehensive approach that integrates the YOLOv8 object detection framework with Long Short-Term Memory (LSTM) networks to enhance real-time threat detection capabilities. The UCF-Crime dataset, which encompasses a variety of crime-related scenarios, was utilized to evaluate the proposed model's effectiveness. Traditional methods often struggle with accurately detecting objects and understanding complex behavioral patterns in dynamic environments. The existing methods primarily rely on conventional object detection techniques that lack the ability to capture temporal dependencies, leading to limitations in anomaly detection. In contrast, our proposed YOLOv8-LSTM model combines fast and accurate object detection with advanced sequential analysis, enabling the identification of unusual activities such as loitering or aggressive behavior over time. Performance metrics, including precision, recall, and accuracy, were employed to assess both object detection and anomaly recognition capabilities, demonstrating significant improvements in real-time threat detection. The results indicate that the integrated model not only enhances the reliability of surveillance systems but also opens pathways for future enhancements, such as incorporating multimodal data and optimizing the model for edge device deployment. This research highlights the potential of advanced deep learning techniques to improve security measures in public spaces, ultimately contributing to safer communities. The successful integration of YOLOv8 and LSTM marks a significant advancement in the development of intelligent surveillance systems, addressing modern security challenges effectively.