We’re happy to announce another new function – coloring of trajectories according to O/D matrix. This feature completes the O/D coloring theme: You can now select to differentiate by color both the gate pairs’ highlights and individual trajectories.
In other news, DroneDeploy has recently released a report titled Commercial Drone Industry Trends. Predictably, the growth is exponential. Their data is primarily for their business, ortophoto maps – but the growth is spilling into other sectors.
Both the PTW driver and the driver of the vehicle being overtaken are assumed to be rational decision-makers that develop strategies, while commuting in urban environment, trying to get the best outcome for their decisions. These strategies may be cooperative or not based on both drivers’ choices with respect to the distances and safety gaps they leave from the lead vehicle. Since PTW follow unorthodox trajectories, especially in an urban arterial, having detailed naturalistic data is essential, and this is how DFS comes in handy!
We are always looking for partners from academic sphere. Are you a researcher in traffic or civil engineering? Don’t hesitate to contact us!
This video is an excerpt with no motorcycles present. You can see some new features of the Viewer: a lateral acceleration heatmap, distance measurement, and a testing bounding box makes a guest appearance too. Heatmap overlay is a completely new feature: When activated, you can configure it to display a scalar value. You can currently select from various speed, count and acceleration types of quantity. The color map and its value range are configurable as well.
Barmpounakis, Emmanouil N., Eleni I. Vlahogianni, and John C. Golias. “A Game Theoretic Approach to Powered Two Wheelers Overtaking Phenomena.” Transportation Research Board 94th Annual Meeting. No. 15-1425. 2015.
Barmpounakis, E.N., Vlahogianni, E.I., & Golias, J.C. (2016). Intelligent Transportation Systems and Powered Two Wheelers Traffic, IEEE Transactions on Intelligent Transportation Systems, 17(4), 908-916. http://doi.org/10.1109/TITS.2015.2497406
Barmpounakis, E. N., E. I. Vlahogianni, and J. C. Golias. “Vision-based multivariate statistical modeling for powered two-wheelers maneuverability during overtaking in urban arterials.” Transportation Letters: The International Journal of Transportation Research (2015): 1942787515Y-0000000020
We would like to announce a new great feature – reports that can be opened in the most widespread spreadsheet editor of all time (so far). Yes, Viewer can now export a complete Excel sheet listing all the possible things you want to know, starting with turning movements and ending with with an image of the annotations. If you do not use Excel, no reason to be sad – the reports are completely readable in LibreOffice, too, and probably in most other spreadsheet editors.
As an example, here is a video from EXPO Milano. Not exactly from the exhibition though, rather from a nearby roundabout during afternoon. This is another video shot from a building, this time at a very poor angle – one of the legs is completely hidden. You can see easily that the roundabout is severely congested in the first few minutes of video. Looks like an overhaul might be needed…
The video included here is a short example, the total recording includes over sixteen hundred vehicles. Anyway, you are invited to take a look at what an excel report from such a place looks like:
We wish to share a couple of interesting features that will be coming to DataFromSky soon. Currently, DataFromSky understands vehicles only as points. This is an outcome of using the particular technology which we chose. However, this results in one very important drawback: We can’t tell real distance between vehicles. Incidentally, we hope for many applications, which would need precise measurement of distances starting from the actual metal, not an ideal point. After this introduction, it should come as no surprise that we are working on detection of vehicle bounding boxes. This will enable much better distance measurement for safety evaluations.
Another improvement in the pipeline is addition of anonymization functions. As you can imagine, aerial recording in urban areas results in a lot of areas being exposed, while they are assumed to be hidden from sight. To protect privacy, DataFromSky will allow marking of areas for blurring or blanking.
We hope to add this soon, so stay tuned for new versions!
We have a new version of Viewer! You can download it at the home page, or use this direct link. The list of differences includes a lot of things, so let’s take a look at them one by one.
Gate angular sensitivity
Gates can count objects, and filter them based on their type. When we had to cope with chaos on the road in both directions, we added directionality. But sometimes all objects come chaotically from all directions, such as pedestrians. What then?
We now added a new setting for gates, where it is possible to select the angles under which objects must come. If their direction falls outside this angular range, the gate will not register them. With this functionality it is possible to handle yet another level of chaos and uncertainty, should you chance to record such behaviour.
On the left is a mock picture of the concept. Both gates have directionality set. The entry gate (15) has angular sensitivity of 90 degrees and the exit gate (7) has angular sensitivity of 15 degrees. Without directionality, the cones that depict the angle ranges would be on both sides. Note that angle of 90° means a right angle on both left and right: A bi-directional gate with angle of 90° (both default values) in fact accepts 180° on both sides, or 360°, and does not discriminate at all.
Now, DataFromSky Viewer users can measure exact distances between two or more points – even if there’s no road. This function is particularly helpful for a traffic engineer. To start using the new distance measuring tool, right click on any point on the image, and select “add point”. Then define another point on the image in the same way, and you will see the distance between them.
The precision depends on the quality of geo-registration procedure, but it is often better than 40 centimeters. The distances are showed in meters. You can also add multiple points to measure distances between them. If you make a mistake, select the point and move it on the right position.
Vehicle center projection
This is not a feature of Viewer, but yet another good news, so here we go! We silently added this in the last months and forgot to mention it. We now project vehicle definition points onto ground. Typically, they lie in the visual center, but not actual vehicle center. This may not seem like much of a difference, but we missed somewhere between 0.2 to 1 meter that way! More importantly, this error is not homogeneous in any practical way, only for vehicles of the same height at the same place. Given that we meticulously correct for perspective and other factors, losing so much precision was not an option – particularly now that we started systematic work on precision and error quantification and improvement.
Improved load and playback speed
Speed is something that one expects naturally from modern computers and software. This is not always possible with lots of visual data, as in our case, but not always impossible either! Now that we are working on (and with) Viewer in earnest, we focused on speed as well. Believe it or not, but sending numbers to show in the old vehicle list turned out to be a drag. Another thing that ended up significantly faster was drawing all the lines, shapes, notes, boxes and texts on top of the video. Apart from this we changed a lot how the Viewer handles video under the hood.
We are confident that you will like how speedy the Viewer turned out, compared to how it was before!
Apart from the visible things, many bugs were fixed.