Pix4Dcapture with ANAFI
- Thread starter IFRINIMC
- Start date
The Anafi focal length is a fixed one (4 mm).
Pix4DCapture is made by a software company that is part of the same financial group as Parrot, so the firmware updates and the lens hardware is well known to them and there is a ‘Parrot AnafI’´Preset in Pix4DCapture.
This Preset includes the focal length, the resolution of the sensor, the sensor’s dimensions.. and it works really great on its own.
That allows to set the overlap of the missions depending on the heigth for exemple.
The image type produced by Pix4DCapture software is now a rectified JPEG (easier to implement with better results in Photogrammetry softwares) if I understood well the release notes for Pix4DCapture.
Just be very careful to set the gimbal angle manually (if you change it manually, it will keep its position even if you programm a ‘Nadir’ mission...
Pix4DCapture is made by a software company that is part of the same financial group as Parrot, so the firmware updates and the lens hardware is well known to them and there is a ‘Parrot AnafI’´Preset in Pix4DCapture.
This Preset includes the focal length, the resolution of the sensor, the sensor’s dimensions.. and it works really great on its own.
That allows to set the overlap of the missions depending on the heigth for exemple.
The image type produced by Pix4DCapture software is now a rectified JPEG (easier to implement with better results in Photogrammetry softwares) if I understood well the release notes for Pix4DCapture.
Just be very careful to set the gimbal angle manually (if you change it manually, it will keep its position even if you programm a ‘Nadir’ mission...
I did the following video with the Anafi only, part of it was made with Pix4DCapture (the 180 pictures at Nadir) :
Thank You Phil. I now understand the Pix4Dcapture Anafi presets for optimum nadir imagery for mapping. I will experiment with altitude for better GSD. As a new drone pilot I appreciate how Pix4Dcapture, FreeFlight6 and Anafi work together. My first mapping flight was without incident.
Congratulations for your first successful mission !
I also use an slight angle (70° instead of 90°) with Pix4DCapture when I need to capture a terrain with a little slope.
Also use Pix4DCapture at various angles for ‘high models’. 70° at the top then decrease down to 30° when I circle around an object.
The idea is to capture the minimum of sky in the pictures.
I also use an slight angle (70° instead of 90°) with Pix4DCapture when I need to capture a terrain with a little slope.
Also use Pix4DCapture at various angles for ‘high models’. 70° at the top then decrease down to 30° when I circle around an object.
The idea is to capture the minimum of sky in the pictures.
Pix4DCapture just knows the focal distance of the Anafi from the start because you chose the Anafi as the drone model for the mission. So it is not really ‘Preset’ it as it is fixed.
For the gimbal settings, you normally set it in Pix4DCapture and it should stay the same for the whole mission, but I remarked you can change it manually with the SkyController trigger during the mission, without aborting the mission.
It’s better to be aware of that, because if you press the ‘gimbal to the horizon 0°’ by mistake like I did once... all the pictures will be horizontal... and you must re-fly the whole thing as pictures are unusable ;o)
Be also extra cautious with the Ground Sampling Distance... I wanted a GSD around 1 cm per pixel, and I flew the mission too low and finished in a tree (the mission height is the same even if there is a slope !).
Took me some efforts to build a ‘Drone Bazooka’ to recover it from more than 20 m ! (There’s a thread here about that).
I found a way to program a mission to follow the slope now, but I still have to test it. Once again, there’s another thread here about missions that follow terrain variations with the Anafi.
For the gimbal settings, you normally set it in Pix4DCapture and it should stay the same for the whole mission, but I remarked you can change it manually with the SkyController trigger during the mission, without aborting the mission.
It’s better to be aware of that, because if you press the ‘gimbal to the horizon 0°’ by mistake like I did once... all the pictures will be horizontal... and you must re-fly the whole thing as pictures are unusable ;o)
Be also extra cautious with the Ground Sampling Distance... I wanted a GSD around 1 cm per pixel, and I flew the mission too low and finished in a tree (the mission height is the same even if there is a slope !).
Took me some efforts to build a ‘Drone Bazooka’ to recover it from more than 20 m ! (There’s a thread here about that).
I found a way to program a mission to follow the slope now, but I still have to test it. Once again, there’s another thread here about missions that follow terrain variations with the Anafi.
Last edited:
Just started to look at Pix4dCapture, which I only want to use for modelling some of our training area elements. Just installed the application on my P30 Pro and looks simple enough, but as a newbie to Pix4dcapture, I have some questions that I am hoping to get some guidance on for best results.
selecting "Circular Mission" and Anafi in the Settings button I have:
Those seem to be it as far as setting go. Is there anything else I should do/change?
I assume captured images area saved to my device or the SD Card, could I then use something like Microsoft ICE to put it all together, or am I stuck with Pix4dCapture? I am just conscious that for the hobbyist, Pix4dCapture will work out expensive.
Any advice in order to achieve good results would be appreciated.
Thanks
selecting "Circular Mission" and Anafi in the Settings button I have:
- Speed (currently set at about the 35% position on the slider)
- Capture Angle (set at about 25% position on the slider)
Those seem to be it as far as setting go. Is there anything else I should do/change?
I assume captured images area saved to my device or the SD Card, could I then use something like Microsoft ICE to put it all together, or am I stuck with Pix4dCapture? I am just conscious that for the hobbyist, Pix4dCapture will work out expensive.
Any advice in order to achieve good results would be appreciated.
Thanks
Jagerbomb52
Well-known member
I only played around with it just for fun. Really have no use for it other then entertainment. The ones I have done when you set the height the program pretty much tells you if you have enough overlap. Most of mine have been done around 200 feet with a angle set to 70%. Now of course the more photos and angles you get I would assume the better the results but also the processing time will be a lot longer. As for Microsoft Ice no you will need a 3D rendering program. I believe there is a few out there that are relatively free to use but not very good. I use Agisoft Photoscan Professional. Not free and very expensive if you buy it. Many places to get a free copy to try out without paying if you catch my drift.
The other thing is 3D rendering is very taxing on a computer. If yours is not up to the job be prepared for a long day going through the work flow.
The other thing is 3D rendering is very taxing on a computer. If yours is not up to the job be prepared for a long day going through the work flow.
It will all depend on the subject and the final accuracy (here GSD : Ground Sampling Distance) you need for the final result.
The main purpose of photogrammetry with Pix4DCaputure is either to make a realistic 3D model of the subject, or to make a map that’s accurate. For example, the use of ICE will give interesting and ‘nice’ results, but absolutely not accurate.
So the use of specific software is mandatory to ‘build the scene in 3D’ then ‘project it in 2D’ if it’s for measurement purposes.
Pix4DCapture is free. All the images captured by the Anafi are stored on the SD card. Just do not upload them to their servers to be ‘processed’ because you need a paying account for that.
But it’s perfectly fine to grab the images from the SD card with a PC or Mac then use them with ICE or free Photogrammetry softwares (or the demo versions of professional softwares)
For a circular mission of a 20m object like a tower, I would fly 30 m above in a ‘tight’ circle around the object, with a gimbal angle of 80°, then descend to 20 meters with a wider circle aiming 60° with the gimbal. Then a last circle at 10 m with a 45° angle.
If details need to be captured, I would manually take pictures up to 0° angle at the base of the object, trying ‘NOT TO Capture the sky’ (or the less possible sky in the images).
I would then fly a grid mission 30 m above to get info of the terrain at nadir (90° or 80°).
The distance from the Anafi should be always approx the same from the object (never more than twice the distance from image to image) so the algorithms can rebuild the scene.
Avoid moving objects, avoid trees (leaves move, and trees attract the Anafi ;o)
At the beginning, avoid hollow subjects such as electric poles or ... the Eiffel tower, they are tough to rebuild..
Always keep your thumb on the left stick ready to stop the mission by a ‘Fly up’ motion... when Pix4DCapture gets the info, the mission stops and you can fly the Anafi manually.
Overlap is the main word. With a classic stitching software, 40% is considered Ok for overlap. In photogrammetry 80% is the minimum...
Always Start your circular mission From a ‘tooooo high’ position. Try it, guesstimate the circle center point, adjust the mission, then only fly it ... GPS and maps can be surprisingly offset...
The main purpose of photogrammetry with Pix4DCaputure is either to make a realistic 3D model of the subject, or to make a map that’s accurate. For example, the use of ICE will give interesting and ‘nice’ results, but absolutely not accurate.
So the use of specific software is mandatory to ‘build the scene in 3D’ then ‘project it in 2D’ if it’s for measurement purposes.
Pix4DCapture is free. All the images captured by the Anafi are stored on the SD card. Just do not upload them to their servers to be ‘processed’ because you need a paying account for that.
But it’s perfectly fine to grab the images from the SD card with a PC or Mac then use them with ICE or free Photogrammetry softwares (or the demo versions of professional softwares)
For a circular mission of a 20m object like a tower, I would fly 30 m above in a ‘tight’ circle around the object, with a gimbal angle of 80°, then descend to 20 meters with a wider circle aiming 60° with the gimbal. Then a last circle at 10 m with a 45° angle.
If details need to be captured, I would manually take pictures up to 0° angle at the base of the object, trying ‘NOT TO Capture the sky’ (or the less possible sky in the images).
I would then fly a grid mission 30 m above to get info of the terrain at nadir (90° or 80°).
The distance from the Anafi should be always approx the same from the object (never more than twice the distance from image to image) so the algorithms can rebuild the scene.
Avoid moving objects, avoid trees (leaves move, and trees attract the Anafi ;o)
At the beginning, avoid hollow subjects such as electric poles or ... the Eiffel tower, they are tough to rebuild..
Always keep your thumb on the left stick ready to stop the mission by a ‘Fly up’ motion... when Pix4DCapture gets the info, the mission stops and you can fly the Anafi manually.
Overlap is the main word. With a classic stitching software, 40% is considered Ok for overlap. In photogrammetry 80% is the minimum...
Always Start your circular mission From a ‘tooooo high’ position. Try it, guesstimate the circle center point, adjust the mission, then only fly it ... GPS and maps can be surprisingly offset...
Last edited:
This thread has got me interested in trying this out - any pointers for opensource or free trail type software to do the rendering side of things?
Jagerbomb52
Well-known member
This one here is free to try and is pretty easy on the computer for rendering as far as I know.
3DF Zephyr - photogrammetry software - 3d models from photos
3DF Zephyr allows you to automatically reconstruct 3D models from photos. Zephyr is the perfect tool to capture reality: a complete photogrammetry software.
www.3dflow.net
A more complex one to use is ‘MeshRoom’ (Free but still usable for a beginner), and a terribly complex but lots of professional softwares use their Algorithms is ‘MicMac’ made by the French National Institute for Geography (IGN). Be prepared to command lines and not really friendly interface for this last one, but it’s also free.
@Agustine Thanks for the tips 
know exactly where yo are coming from. Computing power is not something I have issues with, I work with some incredible computers that we use for Virtual Reality ..... Say they will be re-purposed in their down time
@MustangPhil Thanks for the tips, I shall be having a look at the options you talked about and put some practice in.
I am looking at modelling some of the static scenario's we use for training firefighters from all around the world, things like our A380 Simulator, Tank Farm, Helideck, they should all make good models.
Now all I need to do is get ATC to ok me to fly in quiet times at the airport.... to be honest not much flying at all going on at the moment.
Thanks guys, I just love this forum.
know exactly where yo are coming from. Computing power is not something I have issues with, I work with some incredible computers that we use for Virtual Reality ..... Say they will be re-purposed in their down time @MustangPhil Thanks for the tips, I shall be having a look at the options you talked about and put some practice in.
I am looking at modelling some of the static scenario's we use for training firefighters from all around the world, things like our A380 Simulator, Tank Farm, Helideck, they should all make good models.
Now all I need to do is get ATC to ok me to fly in quiet times at the airport.... to be honest not much flying at all going on at the moment.
Thanks guys, I just love this forum.
One year ago, you had to start an app made by Parrot (Ctrl+Parrot’), then you could go back and forth between Pix4DCapture and FF6.
Now this app is no longer needed (and does not exist nor supported), and it is not advised to use FF6 when you intend to use Pix4DCapture.
The fact is, if both apps are running simultaneously in the phone or tablet... it leads to troubles !
So when I use Pix4DCapture and need to see the pictures the Anafi is taking, I use the Camera icon in the top of the app, that lets you see what the Anafi is seeing. You can even magnify this icon to see it full screen.
If anything goes wrong during the mission, or if in doubt, just ‘fly up’ using the left thumb joystick. It will avoid troubles ahead and will abort the mission. The commands of the SkyController are then active to fly it back home.
Now this app is no longer needed (and does not exist nor supported), and it is not advised to use FF6 when you intend to use Pix4DCapture.
The fact is, if both apps are running simultaneously in the phone or tablet... it leads to troubles !
So when I use Pix4DCapture and need to see the pictures the Anafi is taking, I use the Camera icon in the top of the app, that lets you see what the Anafi is seeing. You can even magnify this icon to see it full screen.
If anything goes wrong during the mission, or if in doubt, just ‘fly up’ using the left thumb joystick. It will avoid troubles ahead and will abort the mission. The commands of the SkyController are then active to fly it back home.
For two months I have owned my Anafi. I have started FreeFlight6 to confirm connections and good GPS, then I open and start Pix4Dcapture. My next mission I will just use Pix4Dcapture. The same checks are present, just in a different format. I will also test the ‘fly up’ command to become familiar with it. I should have watched the Pix4D tutorial. The first thing said is not to run the manufacturer’s app in the Background.
Last edited:
I have been documenting a construction of a building with steel frame. I use Pix4Dcapture for acquisition and Metashape for processing. The steel frame makes for unsatisfactory models as you said. Any suggestions to rectify the open structure issue in photogrammetry. BTW, I bought Metashape because Pix4D trial didn’t allow processing beyond the mesh, I.e., no ortho.
@IFRINIMC : a steel frame or any ‘hollow shape’ is difficult to work and obtain a satisfactory model.
From a recent experience, I had to model the roof of a company that was made of a steel frame.
I had good results editing the dense point cloud and removing manually points that were ‘inside’ the frame or the ones you can see through.
The process is rather simple but a little long, and the same whatever the software you use :
* Copy-paste only the area of the frame you need to work with or group them in a layer that will allow their selection. Hide the rest of the points to avoid unnecessary selection/removal of interesting points of the scene. For example, select the frame from above making a small rectangular marquee selection.
* Isolate the metallic selected frame points and select then remove the points in the steel frame that are ‘not the steel frame’. They are sometimes easy to find, because they are not the same color as the metal frame.
* Then gather all other points that make the scenery around the steel frame.
* Only launch the mesh generation from this step on to avoid ‘bleeding skies’ or objects through the metallic frame that make a ‘thick white structure’ all around the bars.
It was much better this way, although far from perfect.
This technique is proposed in a video on YouTube made by Pix4D as a tutorial.
Just like Dali said : ‘if you want to sculpt an elephant, take a stone and remove all material that does not resemble an elephant’
From a recent experience, I had to model the roof of a company that was made of a steel frame.
I had good results editing the dense point cloud and removing manually points that were ‘inside’ the frame or the ones you can see through.
The process is rather simple but a little long, and the same whatever the software you use :
* Copy-paste only the area of the frame you need to work with or group them in a layer that will allow their selection. Hide the rest of the points to avoid unnecessary selection/removal of interesting points of the scene. For example, select the frame from above making a small rectangular marquee selection.
* Isolate the metallic selected frame points and select then remove the points in the steel frame that are ‘not the steel frame’. They are sometimes easy to find, because they are not the same color as the metal frame.
* Then gather all other points that make the scenery around the steel frame.
* Only launch the mesh generation from this step on to avoid ‘bleeding skies’ or objects through the metallic frame that make a ‘thick white structure’ all around the bars.
It was much better this way, although far from perfect.
This technique is proposed in a video on YouTube made by Pix4D as a tutorial.
Just like Dali said : ‘if you want to sculpt an elephant, take a stone and remove all material that does not resemble an elephant’
