Yesterday I went over to Pitstone Mill again get some more captures to try and overcome the problem that Gordon mentions here:
Your windmill has the same problem as a stack of wood I recently made a 3D model of, and that is the wrong axis is ‘up’. It makes it impossible to view the object of interest from the side, unless it is sideways, with the ground vertical instead of horizontal.
It seems that when all the images are taken from the side, the assumption in the software is that the up axis pokes straight out the front (of the screen), so you can rotate it about that axis, but not about the axis vertically up the screen.
I think that if you fly a lawnmower pattern over the top of the windmill with the camera at -75° and say 85% overlap, then the real up should actually be up in the model, and not sideways, and the rendering of the blades should also improve, assuming they haven’t moved!
When i got home I loaded the original images & the new images into a project and processed them using these settings:
The folder 3D Model photos contains the original photos of the windmill and model files 26-08-22 contains the mapping mission photos from yesterday, any help in sorting this greatly appreciated
I think the double layer problem is due to the 30m difference in the EXIF between the 2 sets of photos, I just used part of the dataset here for a test.
I wonder how that came about, photos taken with same drone, tablet etc and from virtually same launch point, and both set at AGL heights, cheers for looking👍
After just under 2 hours I checked it and it said it couldn’t process the data set, I hit resume and after several minutes this is what it kicked out, the camera circled in red as can be seen from the thumbnail top right should; actually be round the other side somewhere
I would avoid the circular facade - the wall of cameras, I think there is too much background being included. You could also include a geo.json boundary file to help with that.
I’d go with something more like this pattern (although I had trees and power poles to avoid), and keep the distant background out of view
It’s probably best not to have people moving around on it if possible too (I realise that may be difficult), as they are changing positions between images, potentially obscuring features.
Cheers Gordon, so how did you actually fly the lower part of that, I’m very new to all this, probably trying to run before i can walk, and what exactly is a geo.json boundary file, and how would i include it?
I flew all the lower camera positions manually, stopping for each image, aiming to include insides of the old house through the various holes where possible. You should be able to do similar to get views of the back of the blades on the windmill- but don’t get too close, and be sure there are enough features included in overlapping images for them to be tied together. Aim to minimise distant backgrounds if you can.
You can draw a boundary around the object of interest, there are a few ways to do it, but I find this site the easiest:
Draw and save a polygon boundary say 25-30m out from the mill. You’ll need to rename it from xxx.geojson to xxx.json and load it in the options - boundary: xxx.json.
Save it in the same directory where you have to images, as that’s where it will initially look for it.
