As you can see on some of the building faces, the model is experiencing a handful of warping and missing data issues.
I flew the following missions -
150’ altitude, 45 degree gimbal, crosshatch 3D, 80% overlap front and sides
150’ altitude, -90 degree gimbal, crosshatch 3D, 80% overlap front and sides
65’ altitude photo only pass, 0’ FPA, 12 images around the perimeter
150’ altitude photo only pass, 0’ FPA, 10 images around the perimeter
Originally, I was having issues due to (what I thought) was the omission of the oblique images that I now believe that have captured on my perimeter photo passes. Clearly I am still missing some pieces to this puzzle.
If there is any other information I can provide please let me know and I’d be more than happy to.
I’ll mention that I turned off the perimeter 3D because I was shooting the photo passes, and I’ve read many times that you can apparently have too many oblique photos and this causes issues, but it sounds like I may have misunderstood that.
Personally I don’t use perimeter unless contracted to. I only mentioned it because of the image count. It just doesn’t make sense to introduce that many images with the way it does a POI heading instead of perpendicular away from the outer boundary. I have never had luck with that amount of skewed viewpoints.
If the structure is more circular or square it’s ok because you can do a separate flight with a single nadir track and the perimeter approximately 50ft above the structure but when you have a long structure like this you end up with a bunch of images with very acute angles that just introduce unnecessary data. It’s better to track perpendicular to the face of the structure and while you could automate it in something like Litchi I prefer to just manually fly it for safety purposes.
Facade would be bad on this one because it is a clear site and a rectangular structure but when there are trees and equipment in the field we are placing a whole lot of trust in the program with you finger on the pause button hoping that obstacle avoidance does its job. Unfortunately there’s not a lot of drones with good side OA.
Your specs are pretty close but my main theory is to have the camera at a 30-40deg pitch, at or slightly above the roof line and get the entire face in the image. Flying parallel to the face but heading perpendicular. This distance from the face will be determined by these factors. I manually capture the images as well trying to get at least 3 images on every point.
When you get to any really intricate details than are thinner than about 6" try to get 20-30ft from them and at least 3 vantage points. In my mind that would be the nooks and crannies around the towers and the AC units.
Parapets are always tough because they are thin enough and totally subject to the software’s capabilities and settings. This is a downside in cloud processing because we don’t have any control. Yours like pretty good and I suspect with this few additional detail shots they will remain good or maybe a little better.
I’m a little unsure about why I would have less images than you expect on my first flights, even without the perimeter mode engaged.
All of that makes a lot of sense, especially the sections about how the images not perpendicular would be difficult to process since this is not a perfectly square or circular structure…I will manually capture those images around the perimeter staying straight on to each side.
Sounds like my overall plan will be…
Manual perimeter flight at roughly 25’, roughly -35* ensuring perpendicular flight and capturing full facade in each shot.
Manual flight capturing small detail points, again maintaining perpendicular flight.
I see now. I calculated the mass site area as it was shown instead of just the small perimeter at the structure. When you fly over structures you should think about upping the percentages because although you are at 150’ AGL you’re probably only at 120-130ft over the roof so what shows 80% overlap is lower. I would tweak it to add another lane. They really need to get us a locked rectangular option and an offset function would be nice.
I think your plan from the previous post looks good. Are both 150ft flights this same perimeter? I would offset it a little more so on the return the first shot of the bottom of the wall is at the top 1/3rd of the image frame. Looking forward to seeing the next result!
Just moving the perimeter out a little further so when the drone turns back it is getting a good 3 to 4 pictures of the side it’s running to. Some programs allow you to set a turnaround distance beyond the perimeter but obviously DD just stops there.
It is certainly better…But there seems to be an issue with some photos that appear uploaded, but are not being included in the map. Particularly from this corner. The image is first in the list, but isn’t being used. Strange?
It’s strange…I had 341 images in my folder and it shows that there are 341 images uploaded.
However when I click on the images for that corner they don’t take me anywhere in the map. I can even see with that those images are missing from the little ghosted angles that depict where the photos were taken from.
For those interested, I’m STILL waiting on support for this.
I just need to know why the images that are IN the uploaded folder, contain geolocation data, appear clear and usable, are for some reason excluded from the scan resulting in errors in those locations.
Very frustrating that this has not been answered at this point and that I’m on my third support technician.
$500 a month is a lot of money spent to wait nearly 2 weeks for answers.