The team of South Korean researchers has recently announced the development of a new technology capable of detecting and predicting criminal actions using artificial intelligence (AI) and data from video surveillance cameras in real time. The system, named “Dejaview,” expands the traditional role of video cameras, making them not only a tool for observation but also an effective means of preventing crimes.
The Research Institute of Electronics and Telecommunications (ETRI) completed the development of “Dejaview,” an innovative technology using video recordings from observation cameras, crime statistics, and location data to detect potential crimes and predict the likelihood of their occurrence.
The main approach of the Dejaview system is based on the fact that crimes are often repeated according to certain templates, such as type, place, time, and method of committing. The developers focused on identifying these repeated patterns, with the AI system comparing current conditions with data on previous incidents to predict the likelihood of a crime.
The “Dejaview” system is intended for use by the police and control centers of video surveillance cameras nationwide. The AI analyzes real-time camera footage and past crime cases to identify and prevent potential threats like stalking, theft, fires, or drug trafficking. ETRI mentioned that the system enables quick response to situations and helps in tracking suspicious activity.
In collaboration with the SOCHOO-GA district in Seoul, the institute developed a Crime Prediction Map (PCM). By analyzing over 32,000 videos since 2018, the map divides areas into zones based on time, location, and type of incidents to determine high-risk areas for criminal activity. The real-time display of risk levels for each zone helps video surveillance operators concentrate on the most dangerous areas.