Technology

Technology2021-08-27T07:53:29+00:00

In order to create trajectory data sets, we need to detect all road users precisely in the recorded video data and track their movements over time. In order to ensure the highest quality, we use high-resolution 4K video recordings and state-of-the-art computer vision algorithms. Using deep neural networks we can precisely detect and classify every road user in every frame of the recording.

There are many ways to record trajectory data. Each recording method has its own advantages and disadvantages. However, only with the help of a drone is it possible to achieve the highest recording quality without influencing the traffic. From a bird’s eye view, road users cannot occlude each other and every road user on an entire road section can be tracked with high accuracy. Due to the high flight altitude of the drone, road users are not aware of it and behave naturally.

Centimeter-Accuracy

Through our highly-optimized algorithms and quality control processes, we achieve accuracies of less than 10 centimeters. Our False Positive and False Negative rates are much below 1%. The accuracy is constant over the complete measurement area.

Complete 360° scenarios without occlusions

From bird´s eye perspective, the complete scenario is measured without occlusions caused by road users. Every road user that is visible from the air will be tracked by our algorithms.

 

Naturalistic unbiased traffic behavior

 

Our data contains completely naturalistic behavior of road users as they are not aware of the measurement.

Frequently Asked Questions

What are the advantages of drone-based capture reference (ground truth) measurements?2021-07-16T19:01:42+00:00
  • All road users in sensor´s field-of-view are getting measured – both instructed and uninstructed road users
  • Centimeter accuracy with no occlusion by other road users
  • Costs are reasonable and do not scale with the number of road users
What are the advantages of drone-based captured trajectory datasets?2021-07-16T18:58:57+00:00
  • Uninfluenced traffic
  • Highly-accurate data
  • No/little occlusion
  • Large coverage area
  • High Efficiency
Which road user types are measured using levelXdata technology?2021-07-16T18:57:23+00:00

All relevant road users are tracked by our algorithms.

Those are i.a.:

  • Car
  • Van
  • Bus
  • Truck
  • Trailer
  • Motorcycle
  • Pedestrian
  • Bicycle
  • Animal
What area or road stretch are you covering?2021-07-16T18:54:23+00:00

In urban areas, we are covering from 80 x 40 meters up to 160 x 80 meters depending on the given situation and the required accuracies.

At highways, we are covering highway stretches of about 420 meters, which is unique due to our large flight height. At some locations, this flight height is not possible and at least 160 meters are covered.

In which format do you provide the scenarios?2021-05-19T07:59:06+00:00

The trajectory datasets are provided in CSV-format.

Our scenario dataset will be provided in OpenScenario and OpenDrive format.

Which accuracy is achieved using the drone measurement?2021-05-19T07:43:14+00:00

We achieve an accuracy of typically below 10cm. Our false positive and false negative rate is below 1%.

Our Research Papers

uniD Dataset

Go to Top