Researchers Develop AI-Based Road Pothole Filtering Program

The Korea Institute of Civil Engineering and Building Technology (KICT) has developed a "road pothole filtering program" that establishes an emergency repair system for frequently occurring potholes on roads. The pothole inspection screen utilizes an artificial intelligence pothole filtering program (blue: pothole, red: non-pothole). Image source: KICT. Potholes are commonly referred to as "landmines on the road" and refer to the phenomenon of road damage, where a portion of the asphalt sinks into a bowl-shaped depression. Potholes occur when a large amount of rainfall infiltrates the road surface, weakening the underlying foundation and causing the asphalt pavement to collapse under the weight of passing vehicles. Due to global warming in recent years, abnormal weather phenomena such as heavy rain and widespread snowfall have increased, leading to an increase in the occurrence of potholes. Potholes formed on roads inconvenience vehicle and pedestrian traffic and can also cause various degrees of traffic accidents. A research team led by Dr. Moonsup Lee at KICT has developed a new pothole filtering program. This server-based program will be added to the Public Data Management System (PDMS) to review the initial pothole information detected by the pothole detection program. The pothole detection program, operated by the Ministry of Land, Infrastructure and Transport's Land Management Bureau, is a mobile application used to detect potholes. However, due to device performance limitations, high-spec programs cannot run on mobile devices. Therefore, it is necessary to implement a system that can filter out objects similar to potholes (such as shadows, lane markings, and tires) for the initially detected pothole information. The research team has developed and trained an artificial intelligence algorithm to exclude objects other than potholes from the initial pothole information transmitted to the server, effectively selecting genuine potholes. Once the training of the pothole filtering program is completed, it will begin inspecting the initial pothole information. The algorithm is designed to use inspected potholes and other objects as training data to further improve the program. Dr. Lee's research team has established a system that automatically transmits validated pothole information to road maintenance personnel at the Ministry of Land, Infrastructure and Transport every three hours using the developed pothole filtering program. Dr. Moonsup Lee, the lead researcher, stated, "We expect that the newly developed system will reduce the processing time for emergency pothole repairs on roads."