Developing world’s first AI system to detect bridge deterioration

Enabling simple camera-based inspection of bridges and other aging infrastructure to facilitate early repair and service-life extension

Tokyo, Japan — Kim Chul-Woo of Kyoto University’s school of Engineering, in collaboration with NTT DOCOMO, INC., announced that they have developed what is believed to be the world’s first artificial intelligence-based system for inspecting bridges based on vehicle weights and structural deflection estimated from video footage.

The system will be deployed on a trial basis at Yatsuo Ohashi bridge in Toyama Prefecture, Japan between 9 December 2019 and 30 September 2020.

The system uses video images to estimate the weights of vehicles traversing a bridge and the resulting structural deflection caused by the vehicles. AI is then used to analyze the data and estimate the bridge’s structural integrity and degradation. The capability to estimate vehicle weights is expected to strengthen degradation estimations.

Since vehicle traffic and construction environments vary from bridge to bridge, performing periodic inspections and accumulating data on each bridge will enable the accuracy of the AI technology to be improved. In the coming era of 5G mobile networks, the system will realize low-latency transmission of high-definition 4K or 8K video data leading to highly accurate and efficient inspections.

In Japan, bridges of between roughly 10 and 30 meters are normally checked visually or with hammering tests. There are other challenges. Inspection quality often varies according to the skill of the technicians. Also, scaffolding costs and other factors are likely to cause the high costs of inspection.

In response, the use of drone images to analyze surface cracks and corrosion has increased in recent years, but assessing structural integrity from drone images alone can be difficult. The demand continues to grow for technologies that enable infrastructure to be inspected efficiently, assessed accurately for degradation and integrity, and repaired promptly, if required.

The project will use the upcoming trial to help develop an affordable and practical method for inspecting bridges. By 2022, a commercial AI-based inspection system is expected to be launched for the benefit of society.

01.jpg

/Public Release. View in full here.