DeepFine and Seoul Metro Launch AI System to Detect Tram Accident Risks in Advance, Pilot on Wirye Line
TECHWORLD ·
✦ AI Summary
DeepFine and Seoul Metro have begun developing an AI-based safety operations platform that can detect accident risks in advance during tram operations and support control center staff and drivers in responding.
The project aims to develop technology that uses AI to analyze the movements of objects such as pedestrians, bicycles, and vehicles around tram tracks and identify dangerous situations in advance. Research began in April and is planned to continue through December 2027.
Field validation will be carried out in stages along the Wirye Lake Park to Wirye History Park Transit Mall section, and DeepFine set targets of more than 95% accuracy in detecting risky objects and situations, output within 0.5 seconds after a dangerous situation occurs, and a capture rate of more than 90% for non-line-of-sight risky objects.
DeepFine and Seoul Metro have begun developing an AI-based safety operations platform that can detect accident risks in advance during tram operations and support control center staff and drivers in responding. DeepFine was selected for the Ministry of Science and ICT and the Institute for Information & Communications Technology Planning & Evaluation's (IITP) "2026 AX Innovation Company Creative Technology Development Project," and DeepFine and Seoul Metro announced on the 20th that they will pursue development of a tram safety operations platform.
The project aims to develop technology that uses AI to analyze the movements of various objects around tram tracks, including pedestrians, bicycles, and vehicles, and identify dangerous situations in advance based on that analysis. Research began in April, and follow-up research is planned through December 2027 after stage evaluations.
As the lead research institution, DeepFine will take charge of core technology development. The company will develop multi-sensor-based risk detection and analysis technology, 3D spatial modeling, a digital twin control system, and driver support technology based on smart glasses.
With its experience in tram operations and rail safety, Seoul Metro will establish risk categories and operating standards and support field validation on the Wirye Line. Through this, it will pursue the development of a safety operations platform that links AI analysis results to control center and driver response support.
The first-year goal is to develop technology that detects hazards around the track based on fused data from multiple sensors, including CCTV and LiDAR. The technology begins by analyzing each object's movement trajectory, speed, and direction of travel to identify behaviors with a high likelihood of causing accidents, such as approaching the track, entering the track, or sudden movements.
The risk level will then be determined by comprehensively considering the detected object's location, distance to the track, movement speed, and TTC (Time to Collision), and the risk rating will be divided into three stages: safe, caution, and danger. It is also designed to trace the movement paths of each object even in situations where pedestrians and vehicles are densely concentrated at the same time, so that object-specific risk situations can be assessed.
The risk information analyzed in this way will be delivered to the control system and drivers. DeepFine plans to provide driving information such as risk warnings and deceleration recommendations through smart glasses, and to help identify risk factors that are difficult to recognize with the driver's field of vision alone by using sensors and AI in advance.
The research will continue in a direction that combines objects' temporal movement and spatial information, and it plans to create a 3D space of actual tram operation sections. In this space, object locations, movement trajectories, and danger zones will be displayed in real time, and increases in congestion, approaching the track, and abnormal movement patterns will be analyzed as warning signs. The follow-up research plan includes simultaneous analysis of object movement over time and spatial information, as well as the establishment of an integrated control system based on a 3D digital twin.
The field demonstration will be carried out in stages along the Wirye Lake Park to Wirye History Park Transit Mall section. The first-year project is to verify the multi-sensor-based risk detection system. In later stages, the company will apply advanced risk analysis technology and the 3D digital twin control system, with additional performance evaluation as a goal.
DeepFine set the performance targets for this process at 95% or higher accuracy in detecting risky objects and situations, output to the control screen and smart glasses within 0.5 seconds after a dangerous situation occurs, and a capture rate of 90% or higher for non-line-of-sight risky objects. The company also plans to pursue commercialization of the platform based on the technology and operational data secured through the Wirye Line pilot.
The company plans to expand the scope of the technology beyond trams to urban rail systems such as subways and light rail outside new tram lines, as well as smart cities and public safety fields. It also plans to conduct a long-term review of applying its spatial recognition, risk assessment, and on-site response support technologies to other industries such as logistics and manufacturing.
Kim Hyun-bae, CEO of DeepFine, said that because various objects such as pedestrians and vehicles move around trams, preemptive risk recognition and prediction technology is important beyond post-incident response. He also said the company will contribute to creating a safe public transportation environment by advancing technologies that link risk recognition, judgment, and response support together with Seoul Metro.
Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408445
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.