Reading RPLS is free for the whole profession. Members post, reply, and get the members-only rooms.
Tim's 3D experiments, check this youtube link.
It may look like a cartoon, but I can measure, model, and design things from this 3D digital model within 1" over 50'. I can export 3D mesh to numerous CAD and Solids softwares. This model is totally legal, the FAA will have to pry my 25' survey rod out of my cold dead hands!
This is a PIX4D model of the front of my house. I'm just testing the software without a drone. Every photo necessary for this model was done with an IPhone 5s from the ground, and timer shots from a custom mount on my 25' survey rod.
This was strictly a weekend test. Procedural capture could get every part of a house from front to back and every nook in between. In all this model used about 75 images.
It's loosely referenced to WGS84 via iphone GNSS and orientation. If I bought the software (PIX4D) I could kick out a cool version of this model as a KMZ to open in Google Earth. Or I could integrate this data with real survey data surface of the ground conditions for a hybrid model.
Upcoming is a project I'll conduct with Ground Control Points. This will be tightly registered to NAD83 with tons of non-model-added survey grade points for use in ground truthing the PIX4D model.
Very Cool! I hope to see the results of your testing with survey points.
Very cool.
So, if you set "targets" with coordinates and use them to orient the photos, will your accuracy improve?
This is really interesting. How much time was spent processing the 75 photos you took to generate that model?
What kind of range could the right camera setup give you?
JA, PLS SoCal
I don't think relative accuracy will improve much, but targets will improve orientation to NAD83. I did read though that GCPs can help individual orientation of photos that maybe couldn't self register.
Higher Megapixel camera would improve range and quality I'd think.
Total processing time on a bare bones 64-bit 4gig ram computer was about 1.5 hours.
Total Tim time was only a few minutes.
I've done some testing with this and you will see improved relative accuracy in the model as a result of ground control points.
For example, if were to take photos of a sidewalk, relatively linear and flat.. I've often seen a bowing or bowl effect as the software builds the model. Ground control points will flatten it out. The software ties the images together and it has just enough slop that you get an arch over a relatively short distance.
PIX4D is a good pick. interesting how you can integrate CRTP, motion, and vertical aerial into the same 3D or 4D model. my pix4d animated walk through is not a smooth a motion as yours, takes some getting used to
CRTP is beginning to get mind blowing with accuracy and capture capabilities. i can see this application surpassing 3D stationary LIDAR scanning soon
Moe, the latest version released late last week has much better 3D mouse control.
Left button + drag = pan
Shift+Left button + drag = orbit
Makes it way easier to lay out a trajectory for video.
Hardline, thanks for the GCP tip.
Flew the capital building early sunday morning with my buddy and his DJI2+ guerilla style. Very short flight with limited coverage but a pretty neat model overall.
It would not be illegal for you to map your own house with a drone for personal use.
True, and that's what I intend to do in the next couple of weeks. Have to GPS a few more targets and run some reflectorless on the house.
CRTP. No UAV, just iphone and survey rod
What is CRTP? Link please?
CRTP. No UAV, just iphone and survey rod
> What is CRTP?
Close-range terrestrial photogrammetry.
CRTP. No UAV, just iphone and survey rod
Ah, gotcha... I always refer to this technique as structure from motion (SfM). While it could be considered photogrammetry I've always thought of that as where the measurements are made directly from the image.
When I was in college, in one of my photogrammetry classes we actually did a project using CRTP principles.
We had a 35mm camera with B&W film. We took 3 pictures of a wall that had some relief to it (it was in the Art department) on which we had marked some control points (felt pen X). We used a total station to establish the positions of the control points and then, once the photos were developed and printed, we were able to "map" the wall, with all of the relief, to a fairly high degree of accuracy. It was pretty cool.
So it makes sense that now, with much better resolution in photography (and no need for developing and printing of images), that one could get a high degree of accuracy using some solid control and a dense network of high resolution images.
