The intersection of the N237 and the N238 at Huis ter Heide is the scene of a relatively large number of accidents. After an accident, it is important to react as quickly as possible, not only to enable the flow of traffic, but also for the safety of all road users. So it is no coincidence that the Province of Utrecht has chosen this intersection to investigate the scope for detection using existing or new observation cameras. In the pilot, which Royal VolkerWessels company Vialis is conducting with Hyrde and Parquery, intelligence is added to the camera images in order to ultimately improve safety. The successful test in late June, in which a large number of volunteers and stakeholders were involved, provides the perfect starting point to further develop the study of incident detection.
In the study, Vialis is deploying smart cameras and smart sensors for the Province of Utrecht, which analyses images to support road inspectors in incident detection. This helps to make indicators of deviant behaviour more reliable. The technology offers additional traffic information options. It generates reliable key figures about cyclists, pedestrians and heavy traffic. “The technology is also applicable to unsignalised intersections and roundabouts”, explains Bert Batenburg contracts manager for the Province of Utrecht. “With the existing technology, it is hard to generate reliable data. Thanks to this new method of gathering data and incident detection, we can adjust policy and make the roads safer.”
Together with Hyrde and Parquery, Vialis has developed an advanced algorithm. Sharp camera images from Axis Communications were used for this purpose. The algorithm generates real-time triggers and gathers data about traffic at the intersection – all in line with the GDPR privacy legislation, because images are converted into codes and supplied in that form.
In order to enable the algorithm to ‘learn’ quickly, a number of situations were simulated – a collision between cars, a car on the cycle path, a stationary vehicle in the central reservation and a stationary vehicle in the middle of the intersection. Every situation must be detected. In addition, a model has been successfully built to identify near accidents. This is valuable information for policymakers to base their decisions on.
Approximately 35 test individuals gathered at the intersection in late June to test several scenarios under the supervision of professional road inspectors (safety first). A group of 20 cyclists from the cyclists’ association rode in groups of different sizes and approached the intersection using different routes. This enabled an accurate evaluation of the system’s ability to count cyclists. Six cars/vans then tested all the triggers by stopping at the roadside, on the cycle path and on the intersection itself. The algorithm and the dashboard were monitored from the control centre.
Based on an initial analysis, the testers were able to conclude that all the triggers had been successfully deployed and that the dashboard is usable. Sufficient data was gathered to further refine the algorithm and complete development. A final conclusion is expected to be ready by the end of the summer.
Another important conclusion is that, in view of the huge quantity of data generated, the GDPR guidelines need to be followed even more closely. Together with the client, additional solutions are being developed to this end.
A unique aspect of what is at first glance a technical project is the high degree of involvement from the different disciplines. Not only from the road users themselves, but also the road inspectors, who are often first at the scene of an accident. They are full members of the project team. That partnership already makes this pilot unique.