3D, VR, 360° and Street View Photographers  |  Real Estate Agents  |  3rd Party Service Providers
Need a 3D/VR/360°/Street View Photographer? 2,940 We Get Around Network Members in 100 Countries! Ask For FREE referral!
We Get Around Network Forum
3D Showcaseproblem

Issues in showcase1113

PieroBortolot private msg quote post Address this user
Hi guys,
we scan a yacht 2 days ago and we have several issues in the showcase
This is the model:


1- the floors 3 and 4 are not aligned with the other floors of the yacht in the floorplan and in the dollhouse (see picture 2, 3, 4)
Do you know why?


2 - see picture 5 and 6. As you see the scan 233 of the cap. app was made on the floor 4 but it is marked as floor 3. Why? And you can see little rings in front of you. Sometimes, but not always, if you click in these rings, this allow to enter directly in the floor 3


3- the banister is not aligned (see picture 1). Why? There is a way to scan a space without incurring in this issue?
I know that this issue is present in other showcase.

I will appreciate your help

Thanks








Post 1 • IP   flag post
vincent private msg quote post Address this user
This is distortion I have had a similar result in a warehouse.
This is what I think happens
the scans align on the tablet based on comparison with 1 previous scan(typically the previous scan). If the new location looks too similar it can align mis-aligned thus throwing off the rest of the scans that you do from that point on.

As for the rings from another level
I experienced this in a chiller plant where I was capturing from elevated catwalks that were and expanded metal mesh, the stairs from level to level were the same and had open backs. The result was the camera could see through too much and the alignment process was confused when the server processed the file it placed scan rings from different levels on the wrong levels.

It may be best to make a support ticket. My take away from similar experiences is a few things. Pay attention to where the alignment app on the tablet places the scan locations. If you know that you moved 6' and the ring appears on the tablet too close to the last scan location, delete the scan and rescan closer to the previous location. Move in small distances until the scene has enough unique detail to support a longer distance move. The second thing I learned is always start at the lowest elevation and work your way up.
Post 2 • IP   flag post
vincent private msg quote post Address this user
The only thing that I can think of with the banister is to make sure every scan location is level. Could there have been motion in the boat and that caused the instrument move or the tripod to settle while it was scanning.
Post 3 • IP   flag post
andreabortolot private msg quote post Address this user
@vincent
Bannisters: I was present during scan. Boat is 60mts long, no movement at all. We had same issue also in real estate models. And we changed 2 cameras! This is a native hardware problem.. And MP must solve it since this is a lack on a stupid thing
Post 4 • IP   flag post
vincent private msg quote post Address this user
Perhaps the bannister issue has something to do with a change in elevation. I have only seen this happen related to stairs.
Post 5 • IP   flag post
Basic
Member
HarlanHambright private msg quote post Address this user
Closeup objects are going to misalign like that due to parallax I think. The camera is merging image sets from 3 separate lenses and frankly I'm awed by how accurate is usually is. I have window mullions doing the same thing as I got real close.

(The main problem is it's a motor boat, not a sail boat.)
Post 6 • IP   flag post
andreabortolot private msg quote post Address this user
@VINCENT @HarlanHambright
Unfortunately it's not a problem while scanning a stair or because I was scanning a "moving" boat. I (and other users) had same problem on houses and horizontal shapes. This is may become an issue if the client wants to promote an ahestetic place and the 10 milions bannisters decorating his comes out a crap! See attached pic.
Happy Christmas!

Post 7 • IP   flag post
Basic
Member
HarlanHambright private msg quote post Address this user
Again, I've had this problem on numerous occasions and it's complex objects close to the camera. The alignment of the nodal points of the 3 lenses I'm sure are set for infinity.
Buon Natale e Felice Anno nuovo!
Post 8 • IP   flag post
andreabortolot private msg quote post Address this user
Agreed! But I hope MP will solve this issue...they created a masterpece camera and they can't fail onn these details.
Buona Natale a Te!
Post 9 • IP   flag post
vincent private msg quote post Address this user
You know thinking about it. I do scans of mechanical plants
I noticed when I snap into the mesh view to take some quick measurements, I noticed how poorly piping is handled.

I think it has something to do with the nature of the tube shape
maybe somehow the IR projector and sensor get confused by the returned data.
Post 10 • IP   flag post
Basic
Member
HarlanHambright private msg quote post Address this user
I misspoke earlier re: nodal points. It just occurred to me: the reason for the non-alignment of foreground objects is because the 3 lenses' nodal points are NOT aligned. I do a lot of vertical panoramas with the ol' nikon. I usually use a device I invented which keeps the lens in the same spot when you tilt the camera up or down. When I'm lazy, I just use the tripod without it. Photoshop does a good job merging the images when they're off, but the geometry of the images needs adjusting for a perfect match. The 3 MP lenses are a good inch-and-some apart, so there is no way the geometry is going to match, especially on near objects. (At infinity the difference is negligible.) The MP workaround would be for the mesh to be distorted (stretched and bent) so that the top and bottom objects would be concurrent with the center image at the lines of coincidence. I'm not a programmer, but I bet that would be some difficult code to write.
Post 11 • IP   flag post
andreabortolot private msg quote post Address this user
@vincent @HarlanHambright
Thanks a lot! Now I know the trick... Merry Christmas
Post 12 • IP   flag post
36149 12 12
destitute