Wapice has been involved in the development of a neural network-based road damage detection solution for Destia, which will help the company improve road condition monitoring and effectively anticipate the necessary maintenance resources. With the help of diverse data, road damage can be detected faster and more accurately, which means that any damage can be acted upon before it causes harm to road users.
Destia is a Finnish infrastructure and construction service company that builds, maintains and designs not only traffic routes, railways and industrial and traffic environments but also entire milieus. In Finland, Destia is responsible for the maintenance of approximately 45,000 km of the road network.
Neural network speeds up image analysis
The solution designed by Wapice uses the latest available methods to achieve a more accurate, quicker and error-free image processing result than in the case of traditional methods. In addition, the developed neural network-based model can be used in diverse environments without a need to adjust parameters.
– The solution for detection of road damage is based on convolutional neural networks that have an established place in modern image processing. They are particularly used for such challenging detection and classification tasks, says Mickey Shroff from Wapice.
In the example pictures, you can see the original image of the road surface on the left. On the right, you can see the damage and type of damage (colour) found by the neural network-based solution.
When the quality of the input data is accurate enough, the solution performs surprisingly well despite the challenging measurement conditions
Jarkko Rosengren, Measurement Manager, Destia Oy
Benefits achieved
- The solution frees up Destia’s resources for other tasks.
- A faster and more accurate measurement process and the use of digital data improves the cost-effectiveness of road surface analysis.
Article image: ©Destia Oy