I am working on modeling a vehicle and reducing specular distortion, and part of the problem is that ContextCapture is selecting false Automatic Tie Points found in the reflections. I can see which ones are wrong in the Tie Points Navigator screen, but because it is read-only, I can't remove them. I can remove them in the 3D View screen, but it's much more difficult to tell which ones are wrong from that viewer. Is there are way to remove them while looking at the tie points navigator?
Thanks very much,
Looks like an idea.
Official way is to draw masks but it is troublesome and no tools are provided.Vehicles are hard to do you have to use loop sequence so it doesn't find wrong pairs. For vehicles need to prepare them by spraying with some temporary dip. Also using "high" density tie point AT setting could help to remove less accurate tie points.
Also, can you please explain what you mean by loop sequence?
I mean that you take photos in loop or in sequence and then set in aerotriangulation settings to stick to only those pair selection modes, otherwise some photos could be aligned incorrectly. Also I would recommend to do camera pre-calibration and reuse calibrations so reflections cause less alignment issues.Can you post AT report after aerotriangulation?
Sure, here is an example AT of the vehicle by itself, and two different ones where I put painter's tape on the vehicle to try to allow for more accurate automatic tie points. It seems to align just fine, it just seems to create inaccurate tie points on the lower part of the doors where it's seeing the reflection of the rocks on the ground.
Thanks for your help!
You have to get rid of reflection then as it confuses the AT. Maybe limiting viewing distance setting or adding constraints could help. Somehow related to this issue the points in reflection are lower accuracy and CC is aware that they are less reliable but doesn't allow to ignore them.
I agree, it sounds like a similar problem and a good opportunity for a product improvement. I am working on my master's thesis improving vehicle scanning, you have been a great help so I will link it to you when it's finished if you're interested!