Hi , i come across a video surprise me reflective model pic still got a well good model .
https://www.youtube.com/watch?v=KivIw4NW3cc&t=285s
How does he do ? from video, we can saw even using Softbox light , the pictures send to Agisoft are still has some degree of reflection, which as far as i know and experiment these reflective picture can cause photogrammetry fail or bad at construction model , even construct model , but with lots of noise.
However from 4:30 of video , seem the model is constructed perfect .
Does anybody know how he did that so well ? any steps or tricks are welcome.
Well, this is a funny case, because even though the object is metallic, it has lots of patina and micro detail. So once there is an even light in the shooting scene and no harsh directional light, this suits for an ideal object for photogrammetry. If you look closely to where he shoots the subject, it actually could be compared to a rock, with a bit of a shine. So this is not the best case of scanning reflective objects. If otherwise this would be very reflective or featureless, the subject would have to be sprayed, powdered, projected pattern or any other way of defining feature points and matte surface together.
Hi Erik,
from video ( 4:46 ) , you can see all the pictures he shoot has a bit of shiny , but the result mesh show in ( 4:53) is decent .
As far as i know , photogrammetry software can’t tolerate / bear shiny picture , otherwise it will cause noise in finally constructed mesh.
but this mesh ( 4:53) seems got no noise even the constructed pictures has kind of shinny . So , can photogrammetry accept this kind of shinny picture for decent model ?
if so , how should i judge how shinny pictures should be for this kind of decent model ?
I think I have covered clearly why the model looks good above. The shine is not that strong there in the pictures. A shine that would wash out all of the feature points has to really be pure white and very glossy. As mentioned above, what saved this model is all of the tiny detail on the helmet that creates perfect feature points for detection.
OK, Got it . thanks again Erik