INTRODUCING AN AI APPROACH FOR PREDICTING ROBBERY BEHAVIOR POTENTIAL USING INDOOR SECURITY CAMERAS

Abstract

For video surveillance systems to stop incidents and safeguard assets, crime prediction is necessary. In this regard, our paper presents the first artificial intelligence method for indoor camera Robbery Behavior Potential (RBP) identification and prediction. Three detection modules—the head cover, crowd, and loitering detection modules—are the foundation of our approach, which enables us to take prompt action and stop robberies. Using our collected, manually annotated dataset, the YOLOV5 model is retrained for the first two modules. Additionally, we provide a brand-new definition for the DeepSORT-based loitering detection module. A fuzzy inference computer converts expert information into rules before making a final determination about the likelihood of a heist. This is difficult because the thief uses a different technique, the security camera is angled differently, and the video pictures have poor quality.Using actual video surveillance footage, we successfully completed our experiment and obtained an F1-score of 0.537. Thus, we develop threshold value for RBP to assess video pictures as a robbery detection issue and compare experimentally with other similar research. Assuming this, the experimental findings clearly reflect an F1-score of 0.607, indicating that the suggested technique outperforms other robbery detection methods in terms of identifying robberies. We firmly think that by anticipating and averting robbery incidents, the suggested method's implementation might reduce the harm caused by robberies at a security camera control center. However, a human operator's situational awareness improves and additional cameras may be controlled.