No question about it – traffic is a part of civilization, and comes with all kinds of challenges (read: problems). There are over a billion of cars on the planet. The staggering costs associated with traffic jams ($875 million in 2013) and accidents (1.25 million deaths and $518 billion every year) plague both the developed and developing world alike. This is a reason for continuous improvements to the road network.
These changes are best justified and planned with data. The traditional kinds of such data are counts, speed and flow. There are various ways of collecting such data, ranging from observers with notepads to radar, pneumatic tubes and cameras. Video has proven a very useful tool and is currently the fastest growing sector. There is a caveat, though: No camera can see further than the horizon, and no camera can see through solid objects.
Enter the UAV! Cameras can now be carried above all the structures, trees, and vehicles, so that nothing obstructs the view. This new point of view also allows far more than just counting: one can acquire complete trajectories of all the vehicles.
The realization that a UAV-mounted camera can provide valuable data is not entirely new. In 2012, a research project was started at Brno University of Technology. The project linked two teams, one from the Faculty of Civil Engineering, and the other from Faculty of Informatics. The civil engineering team was immediately pleased with the data gathered by UAVs and potential of the method.
That was not the only result, though. The team of computer scientists managed to create something new: a fully automated way of detecting and tracking vehicles in video, algorithmically. Once the video is stabilized and the optics compensated for, computer vision techniques can be applied. After the project ended, the IT team saw that the technology was too good to just be an academic project. A spinoff was born, and given the name DataFromSky.
During the first two years, DataFromSky slowly gained traction. The technique was commercialized as a service for turning aerial video into traffic data. Pilot projects in Czech Republic and Italy showed the potential of the method and attracted interest worldwide, particularly from universities. While the idea is not new, having it actually implemented and available for purchase greatly simplifies matters.
DataFromSky has been presented at a number of trade fairs and events and won a second prize in a Czech startup competition Nastartujtese.cz. Due to its academic roots, DataFromSky is also the subject of a number of academic papers, and indirectly contributed to many more.
DataFromSky does not exist in a vacuum. It is a data processing service and therefore needs other parties to complement it. A UAV operator is needed to actually record the video at the site, which is usually preceded by obtaining permits and working out suitable times and UAV placement. And after the traffic data is extracted from the video, an expert on traffic might be needed to interpret it.
David Herman, the project lead behind DataFromSky and co-founder of RCE Systems says: Practically speaking, DataFromSky created a niche that intersects the fields of traffic analysis, UAV operations and computer vision. We chose to focus on doing what we can do best, and let others’ strengths complement ours: We are experts in computer vision and robotics, but not pilots or traffic analysts. It is natural that we are building a network of partners from these two groups. Since our service is so innovative, we also maintain relationships with universities to stay at the top and pioneer new approaches where possible.
As usual, the biggest hurdle is negotiating the legal landscape to find out whether a survey is possible, and under which conditions.
While DataFromSky was born to help tackle traffic issues, it could do other things in the future. The technology can be applied to monitoring parking lot occupancy, motor racing, cattle counting, and in principle to any moving objects.
The technology itself can also improve further. DataFromSky is a service for offline processing, but a speedup by an order of magnitude would bring it to the realm of real-time monitoring. The DataFromSky team also hopes that UAVs could eventually carry computing modules powerful enough for real-time processing of the video stream.
The biggest hurdle to widespread application is UAV flight time. While an average quad- or hexacopter can manage about 20 minutes, standard traffic survey length is at least an hour. Tethered UAVs are the most promising development in that area, as tethering gives virtually unlimited flight time. Tethering can also change the legal status of a UAV in some countries, which translates to being closer to the road and traffic. However, DataFromSky has already successfully worked with video shot from aerostats, too.
As with everything around UAVs, eventual viability of DataFromSky depends on future legislation and regulations for commercial UAV use. However, Herman remains optimistic: “It might be less easy for us to bid for a project among many operators in the future, due to restrictions or prohibitive costs of operation in urban areas. Still, we are not only in the UAV market, but also traffic analysis market. There are government agencies, as well as established commercial players big enough to negotiate exemptions.”
To learn more, visit datafromsky.com
This article was updated 4/5/16 3:50 pm.
- Handmade Drone Ornament - July 15, 2019
- Data Visualization and Analytics: Q&A with Jan Wouter Kruyt of Propeller Aero - November 7, 2018
- End-to-End Drone Mapping Solutions – Q&A on eBee X and ANAFI Work - October 29, 2018
- 3D Sensing Solutions – Q&A with Cepton Technologies - October 26, 2018
- Would You Rather: Race a Car or Race a Drone - July 18, 2018