Korea Expressway Corporation Invests 1.1 Billion KRW in 'AI Road Pothole Detection'... 85% Error Rate [2024 National Audit]

AI Road Pothole Detection System Developed with Approximately 1.1 Billion Won Budget
Among 517 Detected Damages, 441 Cases Deemed 'No Repair Needed'

A photo of Park Yong-gap, a member of the Democratic Party of Korea. [Photo by Park Yong-gap, Office of the Democratic Party of Korea]

Korea Expressway Corporation introduced an 'AI-based automatic pothole detection system' to quickly repair road potholes, but it was found to be ineffective as errors were detected in 8 out of 10 detections. Concerns about budget waste have also arisen as the equipment, which cost about 1.1 billion KRW to implement, is not fulfilling its role.

According to Park Yong-gap, a member of the National Assembly Land, Infrastructure and Transport Committee from the Democratic Party, an audit conducted by Korea Expressway Corporation in the Gyeongbuk region for one month in March revealed that out of 517 pavement damages detected by AI, 85.3% (441 cases) were errors. Only 14.7%, or 76 cases, were actual damages requiring repair.

Among the remaining 441 cases, 324 were minor damages that did not require urgent repair, and 117 were misdetections such as 'lane paint peeling,' 'road contamination,' 'already repaired areas,' and 'foreign substances.' It was confirmed that the road pavement managers at the branches reclassified the AI-detected data one by one to select those actually needing repair.

In a survey of road pavement managers, only 26% responded that they 'use the AI system,' while 74% said they 'do not use AI.' Seventy percent expressed the opinion that 'system improvement is necessary.' Additionally, 44% stated that 'system improvement is needed to enable real-time confirmation of pavement damage locations and subsequent repair.'

When asked whether the system is well utilized in their work, 40% answered 'no,' while about 20% said 'yes.'

Korea Expressway Corporation stated, "We are conducting research including ongoing advanced studies, error correction, and software deep learning to improve accuracy."

The corporation began pilot application of this detection system in eight branches in 2018 and expanded its use from 2020. Advanced development work continued until last year. They spent 630 million KRW on purchasing AI automatic detection equipment and 469 million KRW on research services for system improvement, totaling 1.099 billion KRW.

Assemblyman Park said, "The equipment procured by Korea Expressway Corporation with a large budget is not performing its role," and added, "It is urgent to improve the system so that prevention, early detection, and timely repair of highway potholes can be properly carried out."