News

Home » News

Two years of DataFromSky

Recently, we celebrated two years since we started commercially! The DataFromSky team was formed shortly after starting RCE Systems as a robotics and computer vision company.

DataFromSky team photo

The decision to start offering traffic analysis using aerial video was a bold move, aiming for international market with a service instead of just contract work for Czech companies. The decision did pay off, though – we feel that at the moment we are the best in traffic analysis using UAVs, worldwide. Over the past two years, we kept growing: You can check the numbers in our earlier post, DataFromSky in numbers.

In case you are now confused about the names: DataFromSky is a service developed and offered internationally by the Czech company RCE Systems s.r.o.. You can check out some of the other things that we did contractually. Of course, only some of the projects are listed. We had the fortune of working with eg. Saab or Lasvit, just to name some.

In any case, we are busy with some new things for DataFromSky. Keep following us for more!

Read more

New: gate pair statistics and track colour by direction

Summer is a time for leisure, vacations and work pace is slower. DataFromSky did not completely sleep, though. Two new functions of the Viewer are now available!

Gate pair statistics

This function is a generalization of origin-destination matrix. For an O/D matrix you need counts. DataFromSky can give you more, though: Instead of just counts, we also offer timing statistics – minimum, maximum, average and median plus standard deviation.

The Viewer got a new dialog for this. It shows a matrix of gates, and tabs for the various measures. When you select a cell, the pair of gates is highlighted as well as the path that vehicles take. If you select more, the highlight colours vary. This way you can easily look at the timings and see what belongs where. If the displays in image overlap, no problem – just drag them where they do not bother you.

Track colour by direction

The second new feature allows you to visually discern what paths vehicles take, too. You can now set individual tracks (trajectories) to be coloured according to entry and exit gate combination. Thus you can easily see where a given vehicle goes, and if you select the non-present tracks to be visible, you can even see the paths all of these vehicles collectively take.

With the direction colouring, this allows easy inspection of eg. lane utilization.

All of the colours can be configured – you can even customize the display to your company brand colour scheme, if you so desire, for images to be included in reports.

We hope you like both of the new additions. We have an endless list of things that can be added or improved, so expect more to come – and keep watching us to learn about more of these exciting developments!

Read more

Dynasim model improvement using our data

Did you know that DataFromSky can provide data for simulation calibration? Citilabs, a traffic modeling software house, recently tried using data captured by DataFromSky to calibrate a model in Cube Dynasim, one of their tools. For the case study, a roundabout in Genestrerio was selected.

The simulation was calibrated based on a comparatively short recording. The resulting observed parameters matched reality reasonably well: flows within 6%, travel time 20%, average speed 16%. This shows that DataFromSky can provide valuable data for simulation, allowing modeling of areas larger than capturable by DFS and contributing greatly to the model prediction quality.

Citilabs have created a presentation which you can browse, as well as a video.

20160519_EN_DataFromSky.pptx

Read more

DataFromSky in Czech Newspaper

Page from Economic Newspaper with finalist medallions Czech Newspaper Hospodářské Noviny (Economic Newspaper) reported about a startup competition Nápad roku (Idea of the year). There were two pages dedicated to the event, with some space allocated to participant profiles. We took part in the competition, too, so the name DataFromSky appeared there as well. The profile reads:

DataFromSky
Founders: David Herman, Aleš Marvan
Subject: Development of mobile robots and machine vision methods
Web: www.datafromsky.com
DataFromSky is an unique tool for analysis of dynamic phenomenons using UAVs, which builds on advanced machine vision algorithms. It offers innovative approaches to interpretation of gathered data about all traffic members in the area of traffic monitoring. We teach drones to see.

Read more

Dynamic heatmaps

We are happy to announce yet another function in DataFromSky Viewer: We have added dynamic mode to heatmaps. Previously, the heatmap displayed data taken from the whole length of video, statically. With dynamic mode, the visualized quantity is taken from an interval around current time.

You can see that there is a lot of potential in having this visualization. The current selection of quantities isn’t everything, of course. We are planning to add various other heatmap options like density, flow, time to collision and others. Follow our news to learn more when these are ready!

In other news, the FAA of USA today finally managed to produce regulations for commercial drone use, ending the legal vacuum or somewhat gray area. Hopefully, this will serve as an example to other countries when deciding how to regulate drones. The focus on safety and privacy is something very understandable. We’re also happy to note that the requirements are very reasonable and operations around road traffic are not significantly hampered!

Read more

DataFromSky participated at Czech Special Forces conference

The Czech Special Forces held a conference dedicated to the emerging drone threats, industry and technologies “Využití UAS a obrana proti nim” (Use of UAS and defense against them). The commercial segment was co-organized by Telink, our partner. DataFromSky took part in this event, too.

You can probably guess that being able to track objects in video is a component of many weapon technologies. In fact, our team has some experience with defense- and police-related projects. We presented DataFromSky as well as what it could do in other domains. Among the considered uses were monitoring of assets during security missions – soldiers and vehicles – and a mobile monitoring set consisting of a van carrying all the required equipment to launch a drone and process the video.

Read more

DataFromSky in numbers

Have you ever wondered how much data passes through DataFromSky? Are you interested in how many objects have been tracked and what distances are involved? We gathered some statistics for you! Here are the numbers for last three months:

35 234 546

meters of trajectories

391 494

tracked objects

230

hours of video

8

countries

2 251 090 500

detector hits

1 957 470

total length of tracked vehicles in meters

87.92%

portion of equator traveled by tracked vehicles

736

kWh burned for computations

What about the software itself?

DataFromSky is written in C++ language and, of course, based on OpenCV library and Qt Toolkit for frontend. The backend detection and tracking algorithms are accelerated by CUDA technology. Here are the numbers:

> 4 000 000

vehicle picture samples

180 000

lines of code

5 500

functions

405

classes

All of these numbers are growing! We are processing new videos every day, teaching the system on more examples of vehicles. We continuously improve our models for position calculation as well.

PS: In the near future, you can expect significant improvement in stitching of videos from more drones.

Read more

DataFromSky presented at SUMO User Conference 2016

For the fourth consecutive year the DLR (Institute of Transportation Systems) held at its headquarters in Berlin the conference of SUMO (Simulation of Urban MObility). It is an open-source simulation platform implemented in C++ which allows modelling intermodal traffic systems including road vehicles, public transport and pedestrians.

Our poster in Berlin with Andrea Marella (right) and Eng. Bonfanti (left)

The flexibility and the continuous development of computational models and new system features make this platform suitable for integration with other applications. Information sharing and the presentation of new usage scenarios involving this platform are some of the main reasons of its rapid development and growing success.

This year’s conference, that took place on 23-25 May 2016, was attended by numerous technicians of the transport field coming from European and American countries. DataFromSky was present, too, in order to expose a project being pursued in cooperation with the Italian partner Andrea Marella, entitled ‘Implementing traffic simulation model with aerial traffic surveys’. You can download the poster here:

pdf Poster_SUMO_DFS.pdf

The aim is to exploit the potential of DataFromSky, in terms of quantity and quality of gathered data, to improve and come up to a more realistic simulation scenario with SUMO. More specifically, the study has focused on the collection of dynamic data of passing vehicles on the survey stretch: this has allowed us to more accurately define the parameters that influence the driving behaviour of the users in that specific environment. The obtained characteristics are then used to define the simulation parameters for the car-following model, and then highlight the huge improvements in terms of virtual representation of the actual observed traffic behaviour.

Training course in Trieste

Andrea Marella talking at seminar in Trieste

Audience at seminar in Trieste

As we announced a few weeks ago, an interesting training course entitled “Methods and techniques for the monitoring and management of road traffic” presented by Eng. Andrea Marella (Trafficlab) was recently held in Trieste. The organization of the event was made possible thanks to the collaboration with the University of Trieste and the Order of Engineers of Trieste.

The meeting was attended by about 70 participants, road and traffic engineers most of all, who have joined with active interest and participation. During the 4 hours course have been described the innovative tools that are currently developing in recent years for the gathering of traffic data. Finally, the opportunity of the meeting has allowed to present a practical case of DataFromSky application, regarding the study of an intersection in Trieste from a video shot from the top of an high-rise. This analysis, already mentioned in previous news was the topic of master’s thesis “The Design of Road Intersection“ of a civil engineering student from University of Trieste.

The great success of the event confirms the interest in this field. Our hope is that other initiatives of this type may in future be promoted by engineering associations and universities, in order to provide an increased awareness of such instruments and of the great potential that they can offer.

Read more

DataFromSky in Czech TV!

logo Cesty k úspěchu DataFromSky and RCE appeared in Czech TV last thursday! We were covered as part of Cesty k úspěchu (Ways to success) – a series about new successful Czech entrepreneurs, currently in 11th season. The series is produced by Prima TV and sponsored by Komerční Banka. Marek Vašut, a popular Czech actor, is doing the interviews.

At the series’ website is a loose transcript of the interview (Google translate). A few parts are missing, though – eg. the funny relationship between founders: Aleš was supposed to teach David at a robotics course, but the actual flow of knowledge was not unidirectional. That’s when the idea of starting RCE Systems appeared first. DataFromSky came later.

As before in Denmark, logos are dutifully blurred by the broadcaster, but you can see our footage and the interviewer openly mentions DataFromSky.

Read more

Drone news – rapidly growing market

chart

Good Drone Lab at University of San Diego published recently an interesting open access “book” titled “Up in the Air: A Global Estimate of Non-Violent Drone Use 2009-2015“. The report is inevitably US-centric due to use of English language news as data sources, but there are several interesting facts listed in the book.

The two distinct questions the researchers asked are: (1) what is the nature of civilian drone use over time, and (2) what regulatory responses exist to use at the international, state, and sub-state levels. Some of the key findings are:

  • The year 2012 was a breakout period that saw non-military use overtake military use.
  • Commercial, emergency services, health and public safety, and environmental conservation sectors are growing.
  • Legislation is sparse and uneven, but growing.
  • Most regulations focuses on controlling weight, altitude, distance, no-fly zones, and operator certification.
  • There is no consensus policy on non-military drone use.

One thing is clear, commercial use is skyrocketing and does not seem to be slowing down any time soon.

pdf Up in the Air_ A Global Estimate of Non-Violent Drone Use 2009-20.pdf

NTIA UAS Privacy Guidelines

Another recent development comes from the US National Telecommunications & Information Administration (NTIA), who arrived at a consensus regarding privacy guidelines for drone use. The resulting material was published under the name “Voluntary Best Practices for UAS Privacy, Transparency, and Accountability“.

The very characteristics that make UAS so promising for commercial and non-commercial uses, including their small size, maneuverability and capacity to carry various kinds of recording or sensory devices, can raise privacy concerns. (…) Operators should use this technology in a responsible, ethical, and respectful way. This should include a commitment to transparency, privacy and accountability.

The published material contains a number of concrete recommendations. The voluntary best practices encourage UAS users to:

  • inform affected persons of UAS use and the collection of data;
  • take care in the collection and storage of information that identifies a particular person;
  • limit the use and sharing of such data;
  • secure data; and
  • monitor and comply with the law as it evolves.

This topic resonates strongly with what we learned about use of aerial video as our business grew, and thus may be of interest to our readers and followers. Some of the privacy concerns are mitigated by the resolution of video that we collect. Arguably, it is impossible to identify people and vehicles since faces and number plates are less than 10 pixels. Of course, combination of the video with other data source is possible.

A more pressing concern is that video from above allows us to see otherwise private places such as walled-in gardens, backyard, roofs, patios… In response to this, we added anonymization functions. We handle videos sent to us with care and use encryption to allow only key team members unrestricted access to complete imagery, when needed during processing.

For the rest, such as informing the people in the area, we rely on the operators collecting video.

pdf voluntary best practices for uas privacy transparency and accountability.pdf

Read more
Page 4 of 12« First...23456...10...Last »