MLIT's Inspection Support Technology Performance Catalog — Japan's National Certification System for AI Bridge-Damage Detection

Japan faces a legal requirement to visually inspect every road bridge and tunnel every five years, with around 40% of the country's bridges projected to be over 50 years old, making manual close-range inspection increasingly costly and labour-intensive.

To give local governments and engineers a vetted, comparable set of AI image-recognition, non-destructive testing and measurement/monitoring technologies as alternatives to manual inspection, with declared accuracy and applicable scope published for each.

MLIT launched the Inspection Support Technology Performance Catalog in February 2019, listing AI systems (such as BMStar) that automatically detect and classify cracks, spalling, rebar exposure, water leakage and lime efflorescence from photographs or drone orthomosaic imagery, alongside non-destructive testing and monitoring technologies. New technologies are added through periodic public announcements, and developer-submitted performance sheets are reviewed against MLIT's standard test items.

Listings grew from 12 bridge technologies in February 2019 to 215 by April 2025 (91 image-measurement, 76 measurement/monitoring, 48 non-destructive testing). Use of catalog-listed technologies became mandatory for specified national-expressway bridge and tunnel inspection items from FY2022, extended to pavement inspections from FY2023, turning a voluntary reference list into an enforced compliance mechanism.

The catalog scaled from a voluntary pilot list to a mandatory nationwide compliance mechanism, but individual technologies' accuracy claims are developer-submitted and ministry-reviewed rather than independently re-tested by a third party, so evidence strength varies by listed technology.