Design and Implementation of a Video Analysis System Using Artificial Intelligence for Real-Time Crime Prevention
Yujin Kim, S. H. Yun, Smith Ann, Seok-Hyun Song, Kihyun Jung · Asia-pacific Journal of Convergent Research Interchange · 2025
This paper presents the design and implementation of an Artificial Intelligence (AI)-based real-time video analysis system utilizing CCTV in public spaces to enhance incident response capabilities against impulsive and violent crimes, including "indiscriminate violence" that has increased since the early 2000s.The system enables effective response through real-time detection of weapons and violent behaviour in public spaces, promptly alerting users to potential threats.It provides user convenience through efficient management of large-scale video data by storing detected incident footage.The research employs two AI models as its core technology: SlowFast and YOLO.The SlowFast model specializes in capturing subtle movements in video for real-time detection of violent behavior.In contrast, in real-time surveillance footage, the YOLO model provides rapid object recognition to identify dangerous objects such as weapons.Combining these two AI models enables more precise detection of weapon possession and violent behaviour, significantly improving the accuracy and reliability of threat detection.Through this integration of technologies, the system explores practical applications in multi-use facilities including, schools, shopping malls, and public spaces, with significant potential for enhancing public safety and crime prevention.Implementation results demonstrate the system's capability to effectively monitor and respond to potential threats in real-time, while efficiently managing storage resources through selective incident recording.This research demonstrates how AI technology advancement can address social challenges, particularly in real-time crime response and prevention, while highlighting the practical implications for public safety infrastructure and law enforcement strategies.The system's dual-model approach represents a significant advancement in automated surveillance technology, offering a promising solution to the growing concerns about public safety in urban environments.The experimental results validate the system's