The system is designed to review primary pothole information detected by the pothole detection program operated by the Land Management Office. This is a mobile-based application used for detecting potholes.

However, due to limitations in device performance, high-specification programmes cannot run on mobile devices. This means that what is flagged as a pothole by the systems is often not so, and instead correlates with shadows, lane markings and tires – a mistake that the new system aims to address.

The solution developed by the KICT team aims to counter this tendency. Using AI tools, the researchers designed a system that can exclude objects other than potholes from the primary pothole information transmitted to the server, effectively selecting the real ones.

Once the training of the pothole filtering program is complete, the algorithm proceeds to inspect the primary pothole information, and then sends the verified pothole information to the road...