Using AI To Spot Potholes

Potholes are the bane of many a road users life, causing damage to vehicles and even to safety.  Fixing them can seem a never ending process for local authorities, but the first step is to be aware of them.  There have been numerous projects to allow the crowd to photograph and identify potholes on their travels, whilst Jaguar Land Rover even tested a system that would automatically detect and report potholes as their vehicles travelled the roads.

A new technology is being developed by researchers at the University of Waterloo to deploy AI to help governments detect potholes and other problems in roads, bridges and even buildings.

“If governments have that information, they can better plan when to repair a particular road and do it at a lower cost,” the team say. “Essentially, it could mean lower taxes for residents.”

Smarter road maintenance

Despite an increase in the number of crowdsourced solutions, today most governments still monitor the roads by driving around and visually inspecting them, with the photos taken then manually inspected.

The Waterloo team have developed an AI-system that can do the assessment automatically, therefore cutting the assessment costs whilst achieving the same levels of accuracy as a human assessor.

“It is more consistent analysis because you’re not introducing the biases of different human beings who look at the data differently,” the team explain.

The system was trained using images captured of roads by Google Street View, before scaling up the training with images from other sources, including companies that work in the field.

The team believe their technology could easily be used to work with media captured by cellphone cameras mounted to vehicles, even as workers go about their usual business rather than specifically looking for potholes.  This data could then be superimposed onto street maps to make it easy to plan and prioritize repairs.

Suffice to say, identifying potholes is but one step in the process, and certainly in the UK a common complaint of people using crowdsourcing platforms is that they identify quite evident potholes but nothing is done because the local government lack the resources to do so.  That’s perhaps a problem too far even for AI technology however.