The technique known as Video-Based Monitoring, in whose development TNO has been instrumental, is able to automatically detect the behaviour of individuals in traffic from video in order to evaluate, for example, the impact of lane narrowing and peak-traffic lanes. Trajectories measured very accurately on the basis of video pictures enable manoeuvres like lane-changing, overtaking and following to be derived and quantified.
TNO has been developing knowledge on traffic and transport for many years. To be able to evaluate the impact of measures like peak-traffic lanes there is increasing need for traffic indicators that have a very high level of detail, such as the speed parameter shown by time or space as opposed to the average of a number of point measurements. Video Based Monitoring offers tools to automatically distil parameters and indicators of individual driving behaviour from pictures with the aim of gaining insight into driving behaviour, for instance. The individual detections together make up vehicle tracks, from which subsequently parameters and functions like speed , lateral position and category can be distilled. Dissecting various parameters can generate interesting traffic analyses in respect of (potential) conflict situations, platooning and overtaking behaviour. Video Based Monitoring thus provides good, useful support to research and recommendations.
The technology uses a detector that recognises and tracks vehicles. Since these characteristics are unique, the toolbox is able in all kinds of conditions (day and night, sunlight, cast shadows, etc.) to very accurately keep the vehicles in view and so we are now able to follow vehicles along an extended trajectory across several cameras and so measure the number of lane changes, for instance. The toolbox comprises a dataviewer that is able to make various dissections of the available information from the video pictures and directly link this information to the video picture. This means that, for example, all the vehicles that overtake on the inside or make improper use of the hard shoulder can be detected and examined. This direct link between data and video pictures enables several issues to be answered quite easily at the same time. The analysis can be manually validated afterwards just as easily.