Preventing Crime and Terrorist Activities with a New Anomaly Detection Approach Based on Outfit

Abstract

Video surveillance systems play an important role in ensuring security indoors and outdoors and detecting suspicious persons due to the increasing violence and terrorist acts every year. In the proposed study, an artificial intelligence-based warning system has been developed, which enables the detection of potential suspects who may carry out criminal or terrorist activities by detecting anomalies in surveillance videos. In this developed system, an abnormality is detected by using the outfits of the people. The YOLOv7 object detection model is trained on our customized data sets, and suspicious person detection is made through outfit information. Especially in cases where biometric data is hidden, dress information makes it easier to obtain information about people. For this reason, the knowledge of outfits is the main point of this study in the detection of suspicious persons. Thanks to this study, security guards will be able to focus on this suspicious person before they pre-empt any crime or terrorist activity. If there are other data confirming the suspicious situation as a result of this follow-up; security personnel will have time to eliminate the crime or attack. The experimental results obtained have been promising in terms of the usability of a person's outfit anomalies to ensure public confidence or avoid risk to human life. Although there are various studies in the literature for the prevention of terrorist or criminal activities; there is no study in which people's outfit is used to identify suspects.