I can't find the video anymore but one guy used some piano wire to make skids for the wingtips on his radian. I think you could use the same idea to protect your servo gimbal from landing rash. He just looped the wire from the front of the wing to the back. It held the wingtips up above the dirt in a landing. One on each side of your gimbal should protect it pretty well.
Photogrammetry
- Thread starter ZoomNBoom
- Start date
I can't find the video anymore but one guy used some piano wire to make skids for the wingtips on his radian. I think you could use the same idea to protect your servo gimbal from landing rash. He just looped the wire from the front of the wing to the back. It held the wingtips up above the dirt in a landing. One on each side of your gimbal should protect it pretty well.
The wingtips of that plane point downwards and there is a skid in the middle. Between that, there is enough clearance. AT least on a nice relatively flat surface with not too long grass heh. But piano wire isnt gonna help much there either.
Anyway, its a disposable gimbal, just one flight to see if it works and how it impacts the result; the next plane is being worked on and will have enough room for a less improvised solution, a large gimbal in the center of the fuselage. I just want to find out if that will be raspberry based or not. Either way, the kibatsu 1 is about to be retired.
Anyway, its a disposable gimbal, just one flight to see if it works and how it impacts the result; the next plane is being worked on and will have enough room for a less improvised solution, a large gimbal in the center of the fuselage. I just want to find out if that will be raspberry based or not. Either way, the kibatsu 1 is about to be retired.
Decided to try and make an orthomap despite the obvious problem. This came out.
Low res (13MP) version, click for full size:
Full res (125MP) here:
https://photosynth.net/embed.aspx?cid=6144773d-6422-4e74-b9b2-bf7b253313fd
If I can solve the jello problem, I think it still shows promise.
Low res (13MP) version, click for full size:
Full res (125MP) here:
https://photosynth.net/embed.aspx?cid=6144773d-6422-4e74-b9b2-bf7b253313fd
If I can solve the jello problem, I think it still shows promise.
The examples here were done with a stock blade 350qx3 combo.
Not the same project but some other examples: https://sketchfab.com/models/5c816d2a9a254a9db13ceca8890519f0
Not the same project but some other examples: https://sketchfab.com/models/5c816d2a9a254a9db13ceca8890519f0
The point cloud is using Pix4D, it is using the video feed from the CGO2 (horrible camera). And I manually fly a grid based on my assessment and gut feeling of the situation also what will likely produce a reasonable cloud. I also have two projects in the works one utilizing GoPro4 black for better imagery and the other utilizing APM and a mirrorless DSLR. For that project it was just shy of 400 video stills, this is a horrible method for data acquisition but it works. An APM controlled versacopter is awaiting development, the plates and other parts in my last order have yet to clear customs.
Cool, you do this professionally ? €6500 or €260/month for that program is a bit rich for my budget, but ill have to give the free version a try to see what pix4d can make of my jello infested, $25 camera pictures. Your results do already prove you dont necessarily have to use very high resolution photo's (your video still would only be 2MP), so that gives me hope 
.
My next UAV for this will have the room for a proper brushless gimbal, then Ill be able to really see what a raspberry camera can (or cant) do and if I can get there with free software.
.My next UAV for this will have the room for a proper brushless gimbal, then Ill be able to really see what a raspberry camera can (or cant) do and if I can get there with free software.
I gave pix4dmapper (free tial) a try. With the (jello free) cell phone images I took of my plane, it does a fantastic job. Far better than visualfsm/meshlab, better than adobe 123 whatever, and I was just using the low quality/fast settings. Sorry for the weird angles, Im still learning:
However, with the less than perfect aerial images, it seems to struggle in the same way that visualfsm does. Using the rapid low quality settings, its far far worse, it fails to calibrate a lot more of the camera's. Ill try the higher quality settings later.
s0berlin, you might want to give visualfsm/meshlab a try with your images, assuming you didnt buy or lease pix4dmapper. Here is an excellent introduction how to use it:
http://wedidstuff.heavyimage.com/index.php/2013/07/12/open-source-photogrammetry-workflow/
Its opensource/free, and thus you're not limited in your export/use of the results. Since I dont yet have a platform yet to make stable geotagged jello free aerial images, I would be very keen to see how yours come out.
However, with the less than perfect aerial images, it seems to struggle in the same way that visualfsm does. Using the rapid low quality settings, its far far worse, it fails to calibrate a lot more of the camera's. Ill try the higher quality settings later.
s0berlin, you might want to give visualfsm/meshlab a try with your images, assuming you didnt buy or lease pix4dmapper. Here is an excellent introduction how to use it:
http://wedidstuff.heavyimage.com/index.php/2013/07/12/open-source-photogrammetry-workflow/
Its opensource/free, and thus you're not limited in your export/use of the results. Since I dont yet have a platform yet to make stable geotagged jello free aerial images, I would be very keen to see how yours come out.
Recalibrated my servo gimbal, stuffed a kitchen sponge behind it (!), and went for another test flight; this time a little higher(120m). Jello is still there, but this is what I got using pix4dmapper.
Still far from perfect, especially close by, but I like where this is going. Not too sure if the improvement is the modifications I did, or the software, Ill rerender it with visualfsm later to compare.
Still far from perfect, especially close by
, but I like where this is going. Not too sure if the improvement is the modifications I did, or the software, Ill rerender it with visualfsm later to compare.
Last edited:
I started using VisualSfM and meshlab. I still use meshlab for some point cloud editing options and re-sampling once I get to that stage. I've tried agisoft photoscan as well, but this is just a much more usable interface and produced the best results for me in a minimum amount of time. I am doing this as a part of a educational research project, so there is an educational price break involved (not huge). But if someone were looking to do some serious mapping and get into more commercial applications this I believe is the way to go. VisualSfM just isn't the same in term or quality and outputs, that being said for open software it is a quite powerful tool.
a
That is a overview of the 3D textured mesh right? What sort of quality in terms of cloud and GSD (ground sample distance) were you getting? Also what is you camera set up?
That is a overview of the 3D textured mesh right? What sort of quality in terms of cloud and GSD (ground sample distance) were you getting? Also what is you camera set up?
Thats using an extremely special and expensive camera!
Well, not really, a $25 3 gram rasperry pi camera module: https://www.raspberrypi.org/products/camera-module/
Its not a really good camera, 5MP resolution, so-so optics, ISO400 is about the limit, but I control everything though the raspberry and can mount 4 camera's (in theory even more) connected to a single raspberry, to do things like NDVI imaging and RGB in a single pass.
Ground resolution is around 3cm/pixel @120m altitude. I cant go much lower because my plane is too fast and the camera wouldnt keep up at least with the stock lens (1-1.5 seconds between photo's) . I used medium settings in pix4Dmapper.
Currently Im mostly being held back by the jello in the photo's. Ive tried a few things, like foam and velcro, but its still present. I dont have room in this plane to make a proper vibration isolation (let alone brushless gimbal) mount and keep the camera out of the airstream. The next one will have all of that, and then I can find out what I can really do with raspberry camera's.
As for photoscan, I tried that too, and got no better results than visualfsm. Only advantage is that it recognized and properly calibrated the image sensor. Once fed those data in to visualfsm, the result was similar. Pix4Dmapper definitely works best for my test data.
Well, not really, a $25 3 gram rasperry pi camera module: https://www.raspberrypi.org/products/camera-module/
Its not a really good camera, 5MP resolution, so-so optics, ISO400 is about the limit, but I control everything though the raspberry and can mount 4 camera's (in theory even more) connected to a single raspberry, to do things like NDVI imaging and RGB in a single pass.
Ground resolution is around 3cm/pixel @120m altitude. I cant go much lower because my plane is too fast and the camera wouldnt keep up at least with the stock lens (1-1.5 seconds between photo's) . I used medium settings in pix4Dmapper.
Currently Im mostly being held back by the jello in the photo's. Ive tried a few things, like foam and velcro, but its still present. I dont have room in this plane to make a proper vibration isolation (let alone brushless gimbal) mount and keep the camera out of the airstream. The next one will have all of that, and then I can find out what I can really do with raspberry camera's.
As for photoscan, I tried that too, and got no better results than visualfsm. Only advantage is that it recognized and properly calibrated the image sensor. Once fed those data in to visualfsm, the result was similar. Pix4Dmapper definitely works best for my test data.
Another quick blurb on some "work" done
These point clouds were just done quick and rushed through Pix4D, needed measurements and a bit of digital information for surface reconstruction.
https://faumfablab.wordpress.com/2015/11/13/site-reconnaissance-reconstruction-2/
These point clouds were just done quick and rushed through Pix4D, needed measurements and a bit of digital information for surface reconstruction.
https://faumfablab.wordpress.com/2015/11/13/site-reconnaissance-reconstruction-2/
Similar threads
