3D Map Worse than Before After Adding More Photos

I 3D mapped a store (Aldis -Short building for reference) looked pretty good, taken from 150ft at -65 Degrees. The only parts that were blurry and distorted was the lower half of the store with the windows, entrance, and cart section. I took some pictures manually the next day up close at around 40 ft high and 40-60ft away from the building and figured the 3D image would get a lot better since everything else looked great. I also added a top down -90 Degree of the roof to get a cleaner edge on the solar panels and figured the program would pick this up easily. Now my image is really distorted and much worse. Any tips for manually capturing, I followed what a youtuber did exactly to get parts that were distorted the first time and his came out great compared to mine.

2 Likes

Were all the pictures taken on the same day?

1 Like

No they were not, I did go roughly around the same time for the second set since it was the same kind of sunny weather the day before.

That could definitely cause it unless you are using GCPs or an RTK/PPK drone. Native GPS on these drones can have up to a 3m shift from day to day.

Ok so its safe to say it should all be shot on the same day, and if something is messed up in post, it may be better to reshoot all together. Question is, could you link a proper YouTube video, if you have the time, that shows how it really should be done? I found this recently uploaded video and I was AMAZED at the detail and perfection Photogrammetry DJI Mini2 - YouTube. It only has 219 views so I hope this man gets the credit he deserves and goes viral in the drone community, but I haven’t found the best tutorials showing how to capture 3D imaging even close to this. Im also not sure why my original 3D Map didnt show a good bottom portion of the store, you think in general I should’ve flown lower than 150ft, maybe 75-100ft? Thank you replying by the way, I and others with similar issues really do appreciate your feedback.

1 Like

Correct, preferably within a couple of hours of each other.

As far as how to do it, if you are looking for a structural model then nadir images will not be that important. I’ll get into this more with an outline of how I would shoot it.

I think the main point of why that model looks so good is because it was processed with reality capture. It would be interesting to know how much clean up and editing was done after. I would guess a couple of hours. Basically saying that this kind of result is not going to happen with DroneDeploy. You could probably contact them to have them reprocess a map with different settings, but the default settings are meant for production and trended more towards terrain. That’s not to say that you can’t get a crisp structural model but everything has to be very well shot and it’s best if these structure is isolated by itself and not a lot of landscape around it.

I can’t really speculate much beyond that without seeing your data, but here is our typical flight planning for a good structural model.

  1. GSD 0.7in/px (determines altitude)
  2. Nadir 75/65 olaps
  3. Enhanced 3D w/ Cross and Perimeter and 70/70 olaps
  4. Manual low perimeter (safely) strafing looking square at the face.

This is one type of flight that DroneDeploy strangely does not have except for the facade which in my opinion is not safe for average users. There’s just too many things that can go wrong and you have to have complete trust in obstacle avoidance. Once you get construction equipment, trees, light poles and power poles, it’s just not worth it.

  1. If there is anything really detailed underneath canopies then I would suggest holding the drone and taking some manual pictures, just walking around.

That’s at least general theory, so you can imagine that different kinds of buildings need different kinds of attention. Comparing that example you shared versus something like a flat roof, industrial building are obviously going to be quite a bit different.

  1. So I’m able to get 0.4in/px at 150-200ft, 200 ft is what is usually puts me at (Not sure if I should now fly higher to get the 0.7 which puts me at 320ft which is a quiet high especially for smaller areas, so I assume Ill jsut keep it at its recommended setting)
  2. Im not sure what Nadir is (Im only using the trial so maybe that’s not an option on Drone Deploy)
  3. I do select 3D w/ Cross and Perimeter at 80/80 olaps
  4. For the Strafing, I did it manually pretty close up but with the building clearly and squared in the shot with manual overlaps in which I can’t calculate, I just estimated to an extent shot by shot when to take the picture

By the way my 2D images came out great and the software is great so far with DD, 3D will just take more practice and understanding

(The link I attached previously, yes he must be using an amazing rendering software that did take hours of editing and he clearly has experience with it.)

1 Like

I should have asked a while ago but what equipment are you using?

  1. The 0.7in/px is a recommendation is to take into account the difference between the ground and the top of a structure. If you are getting 0.7in/px on the ground then you are getting closer to 0.5in/px on a structure that may be 30ft or so tall.
  2. Nadir is just the term a lot of people use for -90deg, looking straight down. Even more confusing some call it -90 and some call it 0…
  3. Check, but I would run a little lower overlaps. Particularly if you run a nadir flight as well.
  4. That’s good. You might try just setting a 2sec interval and try to fly about 5-7mph. I think it depends on how close you can get you might need to fly even slower which can be tough to maintain if like most drones you don’t have a speed control less than normal flight mode. I love that the Air 2S has the Cine mode which really reduces the speed.