The government will use domestically developed neural processing units (NPUs) to convert public closed-circuit television (CCTV) into an artificial intelligence (AI)-based monitoring system.

The Ministry of Science and ICT and the National Information Society Agency (NIA) said on July 28 they will push ahead in earnest with the "Korean-made NPU-based AI CCTV conversion" project.

The project is the first case of large-scale NPU deployment at public CCTV monitoring sites. MSIT will invest a total of 11.68 billion won this year, including 10 billion won in central government funding, to apply NPUs to about 18,100 CCTV cameras.

Gyeongsangnam-do, Ulsan and Korea Expressway Corp were selected as user agencies. Each will introduce Korean-made NPUs developed by Rebellions, DeepX and Mobilint into CCTV control centres.

Gyeongsangnam-do will use the system to quickly detect natural disasters such as wildfires and floods. Ulsan plans to apply it to prevent industrial complex safety accidents and strengthen public security, and Korea Expressway Corp will use it for automatic detection of unexpected situations such as highway traffic accidents and congestion.

If NPUs are introduced into control centre servers, AI models will analyse CCTV video in real time and alert monitoring staff to dangerous situations. MSIT said the system can go beyond detecting whether a risk has occurred by delivering specific conditions in sentences, such as the location of flooding, whether vehicles can enter and whether controls are needed.

MSIT will continue to select user agencies from local governments and public institutions and support AI conversion of public CCTV.

Hong Seong-wan (홍성완), director general for ICT policy at MSIT, said, "I hope NPUs will be widely used for rapid responses at disaster and accident sites," and added, "We will continue to support the AI conversion of public CCTV nationwide."