[Asia Economy Reporter Hwang Junho] In 2022, the year marking the 20th anniversary of the movie 'Minority Report,' the Minority Report concept will be realized in South Korea. A system that analyzes current crime data using artificial intelligence and CCTV to prevent future crimes is being developed by domestic researchers.
The Electronics and Telecommunications Research Institute (ETRI) announced on the 2nd that it will develop the 'Predictive Video Security Core Technology,' which probabilistically predicts what types of crimes may occur by analyzing current CCTV footage, by 2022.
'Minority Report' Approaching Reality
A scene from the movie 'Minority Report.' This film depicts a future where artificial intelligence and big data are used to predict and prevent crimes in advance.
The institute combined the 'future risk assessment system' based on past crime statistics, developed in advanced countries, with 'intelligent CCTV technology.' The institute analyzes how risky the current CCTV situation is by comparing it with past crime patterns. When reviewing CCTV footage where crimes occurred, unusual behaviors are often repeatedly detected, and this system captures such behaviors in current CCTV footage to predict the likelihood of crime occurrence.
For example, if a man and a woman walk a certain distance apart in a high-crime area where four major violent crimes such as assault have occurred during early morning hours, the system alerts that the risk is very high. It recognizes the 'Deja vu' of past crimes to prevent future crimes.
Management with Electronic Bracelet CCTV
The institute improves the accuracy of this system's risk prediction analysis through various data. By analyzing 20,000 court rulings in South Korea, it identifies factors that appear together when crimes occur, and additionally trains the crime prediction analysis model with crime video data from Florida State University in the U.S. and simulated crime scenarios. Visual intelligence technology also identifies attributes such as whether people in the footage are wearing hats, masks, glasses, or carrying items like backpacks.
The institute also manages sex offenders with CCTV. Current location information makes it difficult to determine the intentionality of alarms, and when many people are mixed together, it is hard to identify the target person even by reviewing CCTV footage. The institute uses Person Re-Identification (Person Re-ID) technology to analyze the routes of high-risk individuals such as those wearing electronic bracelets, enabling immediate location of the person via nearby CCTV. This allows for rapid detection and response to high-risk individuals and signs of dangerous behavior.
Predicting Up to 80% of Crimes
In addition, the research team is developing and integrating technologies such as ▲Dynamic Risk Prediction Analysis ▲Human Deep Analysis ▲Active AI Life Risk Analysis ▲Predictive Social Safety Living Lab into the system. They are also establishing response systems including streetlight control, warning sounds, and on-site dispatch, while fundamentally addressing privacy concerns through personal sensitive information protection technologies like video privacy masking.
Once this system is installed in the integrated CCTV control centers of 229 local governments nationwide and police control systems, it is expected that crime risk levels can be monitored in real-time using only CCTV footage.
Kim Geon-woo, Head of the New Authentication and Physical Security Research Lab at ETRI's Information Security Research Division, said, "We will strive to develop a neural network model that enables CCTV not only to detect crimes but also to predict and prevent the possibility of risk occurrence up to 80%, advancing it into a future-oriented advanced social safety system."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
