Using AI to Locate PotholesMarie Donlon | August 03, 2018
University of Waterloo researchers have developed an artificial intelligence (AI) system capable of locating potholes, making the process less expensive and more timely in terms of repair.
The system is designed to analyze images of road conditions captured by vehicle-mounted cameras. Such technology, according to researchers, could help inform workers and government officials about what potholes to repair first. It would also help locate still developing potholes, all in less time than with traditional methods of notification and before the potholes damage vehicles.
Current methods for assessing road conditions are imprecise, usually based on the observations of workers simply driving around. Yet, humans are prone to error and may miss critical, yet imperceptible details concerning cracking and other road surface damage.
As a result, the research team believes the AI system will improve detection and trained the AI on images from Google Street View, in addition to images captured from other sources.
The AI system will analyze the images captured from cameras mounted to a variety of vehicles, including buses and garbage trucks, while those vehicles drive around.
Beyond that, researchers also envision using the AI in combination with drones to inspect buildings, bridges and other structures.