The Electronics and Telecommunications Research Institute (ETRI) announced on July 18 that it has created a system called Dejaview that analyzes CCTV footage, crime statistics, and location data to detect early signs of criminal behavior. (Image courtesy of ETRI)
DAEJEON, Jul. 19 (Korea Bizwire) – In a breakthrough that could revolutionize law enforcement, South Korean researchers have developed artificial intelligence technology capable of predicting criminal activity before it occurs.
The Electronics and Telecommunications Research Institute (ETRI) announced on July 18 that it has created a system called Dejaview that analyzes CCTV footage, crime statistics, and location data to detect early signs of criminal behavior.
Drawing on the observation that crimes often follow patterns in terms of type, method, location, and timing, the research team developed AI technology that compares past and present situations to forecast the likelihood of criminal activity.
The Dejaview technology consists of two main components. The first is a spatiotemporal crime prediction system that assesses risk based on similarities to previous criminal incidents. For example, it would flag a higher risk if patterns resembling past crimes emerge in a secluded area late at night.
Using AI trained on historical crime data, the system analyzes CCTV footage to identify and track situations such as stalking, falls, and the early stages of fires. It can also detect potential drug trafficking or smuggling by recognizing unusual movements of pedestrians and vehicles.
In collaboration with Seoul’s Seocho District, ETRI spent three years analyzing big data from 32,656 CCTV incidents to create a Predictive Crime Map (PCM). This map visualizes areas with a high probability of crime occurrence based on past statistics, categorizing events such as violent crimes, traffic accidents, and fires by time and location.
The PCM’s performance was rated at 82.8% according to testing standards set by the Telecommunications Technology Association of Korea.
In collaboration with Seoul’s Seocho District, ETRI spent three years analyzing big data from 32,656 CCTV incidents to create a Predictive Crime Map (PCM). (Image courtesy of ETRI)
The second component focuses on individual-centered crime prediction, specifically targeting individuals under electronic monitoring who are at high risk of reoffending. While current systems only determine whether monitored individuals violate movement restrictions, the new technology can analyze the risk of recidivism with 95% accuracy when an individual repeatedly violates these restrictions.
The research team completed performance tests using 7,397 daily activity records from electronically monitored individuals.
ETRI plans to implement the Dejaview technology in real-world policing through collaboration with law enforcement agencies and local government CCTV control centers nationwide. The organization also aims to expand its application to AI-based electronic monitoring in cooperation with the Ministry of Justice’s central tracking center to preemptively identify high-risk behaviors among monitored individuals.
The technology is expected to be commercialized by the end of next year.
“This system could be applied to various safety services, such as risk prevention for national infrastructure like airports, energy facilities, and factories, or proactive response to potential threats in personal security scenarios,” said Kim Geon-woo, the lead researcher on the project.
The research, conducted as part of the Ministry of Science and ICT’s project to develop predictive video security technology for proactive risk response based on 5G networks, was selected as one of the top 100 achievements in national R&D projects last year, as well as one of ETRI’s nine representative accomplishments.
Kevin Lee (kevinlee@koreabizwire.com)
