Reputation: 33
the objective of my project is to view a shoe in 3D 360 degrees using point cloud.
So I took four pictures of the shoe, 0 90 180 270 (color images and depth maps).
I put all of them in 3D space and is now able to turn sight of view with mouse.
But then I'm stuck.
How do I put these four sets of individual point clouds into one, how do I fuse/stitch them together to form a complete shoe?
Here's what I'm talking about.
https://drive.google.com/file/d/0B9l2vBy21x6obEFBanVCUk1RYzA/edit?usp=sharing https://drive.google.com/file/d/0B9l2vBy21x6ockdDY1JhUEtCVVU/edit?usp=sharing
I think the main problem is how to identify favorable points from redundant ones.
Any ideas or suggestions?
Upvotes: 0
Views: 5298
Reputation: 1575
You need to find the transforms that rotate the point clouds into the same reference frame. I recommend reading up on spatial rotations and the related linear algebra and geometry. This process is called registration of the point clouds.
So registration involves finding those rotations. The process of I would recommend for doing this registration in the first instance would be guessing an initial transform by eye and then relying on an algorithm like Iterative Closest Point to accurately align. If the alignment is still bad, Moving Least Squares is one relevant algorithm.
More reading about these algorithms can be found here (though the background to the reading is a particular point cloud library): http://pointclouds.org/documentation/tutorials/
Upvotes: 0