We were excited by Pix4D’s Chillon project (http://pix4d.com/chillon/) and decided to make our own 3D reconstruction attempt using our U[g]CS software, DJI Phantom + GoPro and, of course, Pix4Dmapper.
Now we want to share how we made 3D reconstruction of our office.
Pix4D has recommendations for different type of targets. For buildings they recommend to execute two flights around the building with different altitudes and camera angles:
(The image is taken from the Pix4D site)
For a small building it’s possible to take all shots in a single flight, but in most ground station software it’s hard to plan missions where two or more waypoints are located in the same position, but at different altitudes.
Using U[g]CS we can easily create such waypoints thanks to true 3D environment:
So we planned a single mission around our office with two rounds around the building. DJI Phantom 2 autopilot is limited to 16 waypoints, but in this case it wasn’t a problem because we set GoPro to take one shot every half a second.
This is how the mission for our Phantom looks like:
After a very short flight we have got 395 pictures. They were loaded into Pix4D projected and processed by point cloud building algorithm.
First of all, we were impressed by how precisely Pix4D has restored drone trajectory (first picture in this post). More accurately in comparison with GPS track.
And here is the result of point cloud visualization of our office made in Pix4D:
We have also 3D reconstructed another object close to our office — fuel tank station at the airfield:
3D point cloud animation is available for download here.
Taking into account fully automated processing workflow and our zero experience in 3D reconstructions, we think that achieved results are quite good. Next time we will try to get more accurate results by combining different drone trajectories, ground control points and processing modes in Pix4D.
Sincerely, U[g]CS team
Comments
And one more comments about pix4d.
From our experience it's better to use ground control points. 3-4 GCP will give you very high accuracy, much better comparing to geo-referenced images using log files.
Thanks.
If you will have any questions don't hesitate to ask us.
UgCS allows to export telemetry in own XML format, which can be easily converted to any other format you like.
But.
In forthcoming release 2.3 we implemented droneshare.com integration. It will give possibility to download "standard" logs from droneshare.
I hope we will roll out it next week.
Hi - hope this thread is alive! Firstly thanks for creating a much much better ground station for DJI platforms with the 2.4G datalink. I look forward to testing this out ASAP.
With regards to using with pix4d, will we be able to export the data that is collected by your Telemetry Player as a flight log file? I would like to use the data that you have collected to georeference images that my quad has taken with the gopro camera, so that pix4d can use georeferenced images for processing.
Is this possible?
Very interesting. Do you want get something like that?
At my job we're planning to inspect power lines using a thermal camera mounted on a DJI S900, that's where the question came from. Thank you very much for your answer.
For this missions we planned altitude 10 meters above towers with 10 meters distance from axis of power line. We didn't have any problems with drone.
In manual mode we flew much closer (about 5 meters from cables) and didn't have any problems too.
How far did you fly from the power lines? Some say if you fly too near compass and GPS get crazy.
Pictures of reconstructed power line. UgCS used to plan mission, Pix4D for processing.
Hi,
We didn't use georeferenced photos here. Pix4D doesn't require them. We used Ground Control Points to get accurate 3D reconstruction.
Hi.
I have a doubt, how do you get the GoPro's pictures georeferenced?
-
1
-
2
-
3
-
4
of 4 Next