Road traffic counting by drone in Ile de France

We have just entered into a collaboration with CDVIA, a travel and mobility consulting company that carries out, among other things, road traffic studies in order to establish diagnoses and issue useful recommendations for the management and development of road infrastructures.

CDVIA contacted DCOMDRONE as part of an experiment in road traffic counting by drone in order to assess the relevance of this technology, its effectiveness and to establish a comparison with the tools and methods traditionally used.

The drone counting device developed by DCOMDRONE

The experiment that we were to carry out in collaboration with the CDVIA project team was based on the recording, over a period of 2 hours, of a fixed video shot making it possible to visualize the road traffic on a roundabout and its various lanes. access. From the beginning of our exchanges with CDVIA, we considered that the wired solution offering the possibility of maintaining a hovering drone in order to produce a single sequence shot over a long period, without being limited by the autonomy of around twenty minutes of the battery, was not possible due to the height limit imposed by this technology. Indeed, the higher the drone gets, the longer the electric cable that powers it and the more difficult it is to lift. We had calculated that it was necessary to position the drone at a minimum height of 100m to frame the area from the right angle, a height too high to retain the solution of a wired drone.

Why a flight at 100 meters in height?

To allow CDVIA a correct count and therefore a reliable traffic analysis, we must produce an image in which the vehicles (cars, motorcycles, vans, coaches, trucks) are not forced to hide from each other. We must also take into account the vegetation which can hide part of the vehicles during their movement in the roundabout.

Under these conditions, the more the camera is tilted, the greater the risk of masking. Ideally, a drone positioned vertically from the center of the area to be inspected, with a camera inclined at 90°, provides all the necessary guarantees, but this requires that it is legally possible to practice in this way.

The regulations applicable to the professional use of drones prohibit the overflight of roads open to traffic and impose a distance from the ground of 30 meters.

Whenever we have to intervene on a roundabout of 30 meters in radius, we will favor take-off and hovering vertically from the center of the roundabout.

Our feasibility study led us to apply a simple rule: the camera must be tilted at a minimum of 45°. It is therefore imperative that the drone be positioned at a height identical to its distance from the center of the framed area. In other words, 100 meters away imposes a height of 100 meters.

---

To compensate for the impossibility of positioning a drone in hovering flight for 2 hours, we opted for the implementation of two drones, operating one after the other from the same place.

Alternate flight schedule.

When drone no. 1, in flight, drops below the 30% energy reserve bar, drone no. 2 is powered up for takeoff. When drone n°1 drops below 25% energy reserve, the waiting drone takes off to reach the height of 100m by positioning itself 2 meters behind drone n°1 still in action.

The remote pilot is assisted in his maneuver by his control screen on which he observes the image produced by the camera installed at the front of the drone and the telemetry data such as the height of flight, the speed of movement and the meters traveled.

Once in position, he moves his drone 2 meters laterally then 2 meters forward. He then has to tilt the camera downwards to frame the roundabout according to the established model. A "Top Record" is given by remote pilot n°1 to remote pilot n°2. As soon as the second drone is recording, the first can interrupt the recording, begin its descent and land in order to replace its battery. This process is repeated as many times as necessary in order to record the movements of the vehicles over the desired duration.

CDVIA draws up a very positive initial assessment.

• The realization of measurements by drone makes it possible to overcome certain difficulties encountered by conventional sensors. This practice also makes it possible to broaden the field of civilian applications accessible to the drone sector. The counting data from the drones was easy to process and provided a global view of the functioning of the study area.

• The results from the drone videos show relative errors mostly below 10% and overall low absolute errors, while the results from the counting mast videos show relative errors generally between 10% and 25%.

The limitations of the drone system identified can be classified into two categories, the limitations related to the drone technology and those related to the processing technology.

• The first category is mainly represented by the autonomy of the mini drones limited to about twenty minutes, this is conditioned by internal factors, the size, the payload, the type of battery, and external, the weather conditions, the conditions of wind and regulations on the occupation of airspace which can be restrictive and lead to flight interruptions.

• Video processing limitations relate to vehicle detection and tracking errors that occur as a result of partial occlusions, false detections, or due to the variable size of some vehicles in the image and the camera angle. view that does not allow all vehicles to be viewed in the same way.

---

If you want to know more, do not hesitate to contact us. We will be happy to listen to you and advise you.