An AI solution for automatically recognizing and classifying road markings.

The customer

Van Gelder is a contractor, with more than 1100 employees, who realizes above and below ground infrastructure. They are active throughout the Netherlands and beyond. Van Gelder's activities cover the entire infrastructure palette: from area development and road construction to cables, pipelines, networks and installations for controlling and securing road and rail traffic and the transport and distribution of energy, water and data.

The challenge

The primary goal of this project is to automate the manual processing of the images as much as possible. The camera images from the mobile recording systems on Van Gelder's recording vehicles are currently manually processed by an employee at the office. In this process, marks are drawn manually to record a baseline measurement. Then the quality of these markings is examined. The question of Van Gelder is whether the inventory of lines can be carried out automatically.

The solution

To solve Van Gelder's issue, Kaios.ai has developed an artificial intelligence (AI) -based detection and classification model. This AI solution automatically recognizes and classifies road markings. And is able to recognize and classify different types of markings (lines, points and surfaces). This classification is based on the standard of the CROW guidelines. All data is also provided with a geo-location and can therefore be read in via regular GIS programs (.shp).

The model is built using a Computer Vision algorithm in combination with Machine Learning. The expertise of the Van Gelder specialists has been used in developing a good model that meets Van Gelder's wishes and requirements.

The result

Our model largely automates the time-consuming and therefore costly task of manual processing of the camera images. After the automatic inventory of the road markings, the result can be further processed using the Kaios.ai Platform or any other Geo Information program. In the Kaios.ai Platform it is possible to adjust and optimize the data before it is delivered to the end user.