Until now there were very few mobile traffic monitoring systems such as special vans altered for surveillance and monitoring with radar and camera equipment that were quite expensive. Despite them having a very good range the data they provided was limited and useful for a few applications such as speed detection or seatbelt compliance with the speed detection radars serving only a single application. Police, the main user of such systems, also wanted to detect things such as vehicle gap times, aggressive driving, or measure speed with one solution and were looking for a more cost-efficient alternative. Thanks to the combination of drone technology coupled with the latest capabilities of AI traffic analysis a new much more versatile and affordable solution - the TrafficDrone was born.
TrafficDrone mobile traffic monitoring and processing unit is a brand new solution for on-the-spot real-time traffic analysis from a drone video capable of a variety of tasks. It can detect and classify objects into 8 categories and provide full trajectory data for all of them. This enables traffic monitoring, gathering of traffic flow statistics, monitoring a variety of traffic violations including speeding, wrong-way driving, U-turns, wrong-way driving or illegal lane change but also blocking vehicle detection, perimeter breach detection, and crowd monitoring function. Get familiar with the latest innovations in mobile traffic intelligence and learn about fresh solutions bringing together real-time AI video analysis capabilities with drone technology in this webinar.
In this webinar extract, you can learn about:
1. Details of the special equipment needed for real-time traffic analysis - namely the TrafficDrone mobile processing unit and the Safe-T 2 tether station from Elistair (more in the full video by Elistair).
2. The types of data that can be gathered - including traffic statistics, various traffic violations, or perimeter detection alerts.
3. The connectivity, detection, and processing capabilities of the TrafficDrone unit.
Questions and Answers
Q 1. What is the latency of detection of the traffic analysis?
A: With the TrafficDrone unit you can get data with roughly 200ms latency. Therefore, it is ideal for real-time situation awareness and it can directly control action elements such as warning lights or sirens using IO outputs.
Q 2. What traffic data analysis can be done?
A: Traffic analysis unit can detect, classify vehicles and also get their trajectories, speeds, detect whether they are stationary and for how long. You can gather a variety of traffic statistics for multiple directions at once and detect traffic violations such as illegal lane change, wrong-way driving or speeding. You can define a lot of traffic scenarios thanks to the flexibility of the FLOW framework running on the TrafficDrone Unit.
Q 3. How does nighttime traffic data collection and analysis work?
A: Here is and example of our detection at night:
The TrafficDrone unit is able to work in low light conditions meaning that detection still works fine when:
a) the light conditions are either such that you can still clearly distinguish vehicle outlines or
b) there is number of vehicles are clearly illuminating each other
c) there are lamps illuminating the road
Q 4. How is the data accuracy with heavily heterogeneous traffic scenario?
A: The video analytic engine is based on deep neural networks which are trained for multimodal traffic scenarios. Depending on the concrete situation, the accuracy varies from 95 to 100 %. The engine could be optimized for the specific scenes or extended by a new object category and/or attribute if requested.
If you would have a narrow use case – let's say detection of only one type of car then the neural network could be optimized for that and achieve 100% accuracy with a sufficient dataset. The limitations are set by the processing power combined with detecting and tracking multiple categories at the same time.
Q 5. Is this a cloud-based solution and can it run on a private server?
A: DroneTraffic Unit is an all-in-one video-analytic solution. In operation, no cloud or private server is needed. But the system can be deployed on a cloud or private server if requested. To ensure the same video-analytic quality, there must be good data connectivity between the drone and the server/cloud. We also offer TrafficEnterprise which can be used to analyze videos from cameras.
Q 6. What is the accuracy of traffic analytics in ideal conditions at 100 meters height?
A: If we are talking about truly ideal conditions meaning – no occlusions, no atypical vehicles, then for vehicle classification the accuracy will be around 99%, every car will be detected in some way and will be tracked in the whole scene except on the sides (5% margin).
Q 7. What is the value of traffic analytics using drones? The same functionality can be implemented using a stationary camera.
A: There are multiple benefits. The camera loses accuracy with distance while the drone provides accurate coverage of the same and even bigger area compared to static cameras thanks to the top-view. Getting distance related data such as gap times or speeds can be done with very good precision with a drone compared to a camera. And most importantly, it is a super mobile solution deployable in a minute if needed, allowing you to see but not be seen.
Q 8. What is the optimal distance/ height to be able to have good accuracy of counts and classifications of vehicles?
A: The optimal height for traffic analysis with drones with good classification and count data is between 70 and 100 meters.
Q 9. The tethered station can connect with TrafficDrone directly by HDMI only? If yes, is that the tethered station able to work with all drones with HDMI signal for feed-in or still there is another compatibility requirement?
A: As soon as you can connect the video rush from the drone to the DataFromSky Unit through HDMI or SDI, it will be compatible. Tethered stations are compatible with more than 50 commercial drones, to power the connected drone for hours.
Q 10. Is it possible to remote control the TrafficDrone unit for remote config/central monitor purposes if internet is provided on-site?
A: Yes. It is possible to remotely control the TrafficDrone.
Q 11. Can the analysis be done from offline video files?
A: Yes, if you will simulate the live stream but it would take the same time as the video recording. We have a specially designed solution for this use case called TrafficSurvey. It is an easy-to-use online service for processing the video records available at https://ai.datafromsky.com/. If you are interested in an on-premise solution for processing the video records, contact us at firstname.lastname@example.org.
You could also simulate the live stream to do the offline analysis but it would take same time to do as the video recording. Another alternative would be to deploy TrafficSurvey video analysis software from us on the TrafficDrone mobile processing unit.
Q 12. Can the videos be recorded for forensics purposes at the same time we have the traffic analysis in the live feed?
A: Yes. It is possible to record a whole flight or even specific user-defined events. The TrafficDrone unit can be also connected with VMS such as Milestone.
Q 13. How many drones can be connected with the TrafficDrone unit to do a real-time analysis?
A: It is possible to have 2 drones connected to one processing unit.
Q 14. The DFS has a dedicated box that takes the video in from the HDMI or does it work with any video capture device on the PC?
A: The processing unit gets the video data via HDMI with the help of the DJI drone but it can be also connected to an RTSP stream. It features HDMI and SDI input, multiple USB-A and USB-C interfaces, Ethernet port and WiFi.
Q 15. How is the drone connected to the computer for live data analysis?
A: Video feed arrives to the DJI remote control through RF. DJI remote control is then connected to the "Trafic Drone Unit" (DatafromSky) product through HDMI.
• This demonstration was hosted by Elistair
• Learn more about the TrafficDrone mobile processing unit.
• Do you have more questions? Contact us!
• Follow us on Linkedin.
• Visit our homepage to view products and news.
Thank you for reading this article! Have a great day!