**The way Telraam measures speed **

Telraam cannot measure speed directly like a typical police speed gun (a.k.a. a small Doppler radar), so we need to use an indirect method to derive velocities. In an ideal world, if we knew the distance between the camera and the passing objects, or the length of the road covered by the field of view of each individual Telraam unit, then we could calculate speeds from measured travel-times. For example speed could be easily calculated from the time it takes a car to travel along a known length on the road.

Unfortunately each Telraam unit's location is unique, so the distances between each Telraam unit and their host windows are different, and it would not be practicable to ask each user to actually measure this distance between their unit and the middle of the street.

Furthermore, the closer a multi-lane street is to the Telraam unit, the bigger the difference will be between the distances that are present between the Telraam and the individual lanes. As in urban areas building fronts are often only separated by a narrow sidewalk from the road surface, such differences can easily reach up to 50%, meaning that a single distance value (between the Telraam units and, e.g., the middle of the road) would cause significant errors when calculating speeds of objects that pass through the individual lanes.

To make our measurements distance independent, we came up with the following clever solution: Telraam measures the time it takes a car to travel its own length - which does not depend on the distance from which the car is perceived - and assuming the car has the length of a typical car (a value we calculated from the most popular cars sold in recent years in Belgium), we can now calculate its speed. Of course this method has its drawbacks too.

**Reasons why is your speed data seems unrealistic**

There are various factors that influence the accuracy of the speed histogram that is displayed for a given road segment. First of all, a prerequisite to accurate speeds is that the **classification** is accurate. In most cases this is not an issue, but if you see that your Telraam measures a very different split for, e.g., cars and cyclists, then the derived speed histogram will be affected by the misclassified cyclists, as their incorrectly (using a wrong assumed object length) derived speed will be mixed together with the speeds of objects that are actually cars.

Furthermore some scatter is expected even if every car passed in front of the Telraam at an exact 30 km/h pace, as we assume each car has the same **length**, which is of course not the case in reality. There are a few minis, and some big SUVs, that will be further away from the average size. In this case smaller (shorter) than average cars will be "measured" faster than 30 km/h, and larger (longer) than average cars will be "measured" slower. This is also the reason why one should not take speed data from a small sample of cars very seriously, as this effect gets less visible when looking at a large (and therefore more average) sample of cars.

And finally, there are some **outliers** - artefacts - in the data, which can show up as unrealistically high (or low) measurement points in the speed histograms, so if you see a few (%) cars driving 70+ km/h in a 20 km/h zone, that does not mean that there were people actually racing in that street.

To get the most precise idea about the speeds measured in a given street, always look at a large sample of data (so not a single hour with very low traffic levels), and consult the V85 value which is one of the best indicators of the typical speeds, as it is practically unaffected by these spurious outliers.