Seoul to develop AI solution for detection of hecklers at subway stations

By Kim Joo-heon Posted : October 10, 2023, 15:34 Updated : October 11, 2023, 07:56
Yonhap Photo
[Yonhap News Photo]
SEOUL -- Seoul will develop an artificial intelligence-based solution capable of monitoring the movement of people to see if there are any abnormal activities at subway stations. In case of an emergency, the solution will notify a human inspector in full sentences to handle various types of accidents. 
According to the capital city, social unrest is increasing due to the recent incidents targeting an unspecified number of the public. On July 21, a man in his 30s killed one man and injured three people with a knife near Sillim Station in southwestern Seoul. Only 13 days after the stabbing rampage, 14 victims were injured by a man in his 20s who intentionally hit pedestrians with a car and stabbed people at a department store in Seongnam, a southeastern satellite city of Seoul. After such incidents, the Ministry of Justice unveiled its plan to adopt a life sentence that does not grant parole.
According to the capital city government on October 10, Seoul Metro partnered with Seoul Digital Foundation operated by the capital city to set up a model that can detect irregular behaviors at subway stations using AI and big data technology. After the establishment of the generative AI-based solution, the two groups will demonstrate its efficacy at a station on line number 5. After the verification, set to be completed by December 2023, the solution will be expanded to all stations.
The monitoring model will incorporate image captioning, an AI-based process used to generate textual descriptions based on images. Unusual behaviors detected by a closed-circuit television camera would be converted into text. "The abnormal behavior detection solution, based on generative AI, is a technology that can effectively respond to incidents targeting random people such as rampages in the subway," Seoul Metro CEO Baek Ho said in a statement.
기사 이미지 확대 보기