How road condition analysis with computer vision is changing cities and transportation departments
Road infrastructure is among the most important public assets we have. Essentially, modern society doesn’t work without them.
And yet, many cities, counties, and transportation departments struggle with budget and labor limitations that cause their roadways to suffer.
These challenges affect us all. For example, the American Automotive Association estimates that U.S. drivers spend $3 billion a year fixing pothole damage to their vehicles. Naturally, bad road conditions cause public safety problems as well.
Fortunately, machine learning, computer vision, and other technologies have matured so quickly that serious improvements can be made to road maintenance practices — relieving transportation departments of existing cost and labor burdens, improving maintenance, and maintaining higher safety standards.
Vaisala's computer vision team started work on RoadAI in 2015, with the purpose of automating labor-intensive road pavement defect inventories and providing transportation departments with better data for decision-making. The team understood the enormous value in enabling organizations to complete easy surveys with rapid analysis without any manual effort.
The solution was created through close industry partnerships, which enabled Vaisala to stay focused on workers’ real challenges and experiences. One outcome is that with RoadAI, any vehicle or driver can collect excellent video data with little more than a camera and a smartphone app — no need for special education or training.
Other benefits of RoadAI include:
• 4X faster than other methods. The vehicle can operate at normal driving speeds, unlike manual detection, which forces drivers to go very slowly.
• Half the cost of traditional pavement analysis. Organizations can easily do the surveys themselves or inexpensively hire someone to do just the driving. No special survey contractors required.
• Objective, consistent assessments free from human error. Surveys no longer require human assessment or attention.
• Comprehensive data that leads to better, more strategic decision-making. Leaders are able to plan more effectively, and asphalt repairs are more timely and last longer.
How computer vision technology works
Computer vision is a type of machine learning technology, and in simple terms it means using software to interpret an image or a video in same way than human would. RoadAI is optimized to understand pavement deterioration and can identify 11 defect types, which are categorized individually while the tool calculates a pavement condition index (PCI) for each road section.
In following image you can see how detection is done and how defects are quantified. The detection is done in front of the car in a 1 meter (3 feet) deep area to ensure that even the finest and most detailed cracks are captured.
Figure 1: Example from inspector´s vehicle view and computer vision in action
Figure 2: Different defect types’ impact on road condition determination
A new maintenance paradigm
RoadAI is a perfect example of the power of machine learning and automation — and their practical impacts in our lives. Almost any city, county, or transportation department can now monitor network conditions continuously, replace old manual surveys with an automated camera-based system, and reduce human error. Moreover, they can use quality data to make better, more strategic decisions.
Given the challenges faced by cities, counties, and transportation departments today, RoadAI is most likely the only viable methodology on the market for implementing continuous monitoring on road networks.