As part of the testing with photogrammetry, I wanted to start to look at how it works capturing puppets, rather than build a full puppet I decided to use a doll. I blue tac this down to a base and took photos using the same lighting set-up as before.
This test could have worked better. When I put the photos into agisoft it didn't recognise certain areas. Like the legs and arms, and the rest of the model came out really bumpy. I'm not sure if this is because the doll I use has a slight glitter layer on the skin and it was too reflective for the software.
Here is the point cloud
From this point cloud, I create a dense point cloud.
Here is the dense point cloud, at this point, I could see that it was having issues with the legs and arms, but I finished the test to see the results. I cleaned up the cloud and then build the model.
Here is the model solid
The model shaded, you can see on this that the model is all bumpy and has a really uneven surface.
The model mesh
The model confident
And the model textured.
While this test hasn't worked, I have learnt a few things for the next test.
I need to try and get closer when I take photos.
I will find another doll that does not have glitter skin. (I believe this is the reason the arms and legs look so bad as these are the bits with the most glitter, which means as the camera is taking the photos the light will be reflecting all over which agisoft then cannot read.)
I will find a ND filter for my camera, which means that I should be able to get images of shiny surfaces.
I will conduct another test, using the above things and see if I can get better results. I am also going to research how other companies have conducted 3D scans similar to this. I will try to email Matt (old tutor) and see what methods he has used as he has done similar before and captured the wind in the willows puppets.
No comments:
Post a Comment