Home Computer device Uni boffins trains a machine learning model on rescue team data •...

Uni boffins trains a machine learning model on rescue team data • The Register

34
0

South Korea’s capital and largest metropolis has turned to AI for suicide detection and prevention on popular bridges along the important Han River, according to developers of the system at the Institute Seoul Technology Center (SIT).

There are 27 bridges that cross the Han River, also known as the Hangang River, in the Seoul National Capital Region (divided into Seoul, Gyeonggi, Incheon). Many pedestrians in the city walk on it every day. Sadly, counting only suicide attempts from Seoul bridges, the researchers said there are an average of 486 people who attempt to end their own life in Han waters each year. This translates to a large amount of rescue resources needed, which luckily are quite effective at 96%. A more efficient allocation of resources, however, would mean being able to potentially save more lives.

With the current configuration, which according to a SIT document was put into service in February, the bridges are monitored by a set of CCTV screens in a control center, with each small box representing a different part of the bridge. An employee monitors these screens for unusual behavior that may require interference or a rescuer. The collaboration with AI between SIT and the Seoul Fire and Disaster Headquarters aims to help technicians focus better by using machine learning to alert them to scenes most likely to require intervention.

“The objective is to put in place an effective control system and to reduce the number of victims caused by fall accidents”, according to the SIT document, when translated into English via automated language tools.

The collaboration between uni boffins and workers at the Seoul Fire and Disaster HQ has been providing data since 2020, detecting and predicting situations by learning the behavior patterns of those in need of help and posting these images on the controller monitor. These patterns could consist of actions such as pulling on cables detected by sensors. The hope is that in the end the system will learn “the learning of [behaviour] schema of the perpetrator of the attempt “and rescuers have more time to act before the individual acts.

The system is constantly learning, analyzing structured and unstructured data such as report history and call detail text, as well as sensor data, to reduce false alarms and improve accuracy. The characteristics of the bridge and changes in the weather, such as light and wind, are also taken into account.

A video posted to SIT’s Facebook page and YouTube channel shows an engineer operating the surveillance system. In a scene watched by the engineer, the area outside the bridge is grayed out and not sharp, and the movement caused by the wind is stabilized.

Youtube video

Video engineers said the system minimizes surveillance deviation, allowing faster and more accurate response to incidents, saving lives. Currently, the rescue rate is around 96 percent.

Korea has the highest suicide rate across all OECD countries to 23 per 100,000 population between 2016 and 2018. COVID-related difficulties have since increased these numbers and the number of relief shipments on the Han River Bridge, reports show Reuters. ®

TalkNote

To those who read this and have had bad thoughts, click here to find a number to talk to someone. Things will be different tomorrow.

LEAVE A REPLY

Please enter your comment!
Please enter your name here