## Downsampling method: FPS
Core operation: Find some points so that the distance between these points is as far as possible.
What does that mean? As follows:
Find a point (marked with red) first, how to count the point that is farthest from the red point among the remaining black points, and then mark red, and then calculate the point where the remaining black point is far from both red points. Going back is to pick the point furthest from the known point and continue to mark red.
What's the use of this method?
You can remove some very dense points. For example, there is a lot of dense points in one area, but the points are sparse in other areas, so you need to remove some places at this time.
The most common place is deep learning.
## Downsampling method: NSS
Downsampling above the normal vector.
Why do you need this?
First look at the following picture
