Fast search algorithm for point cloud data

Question

I am working on Zed-mini camera with Nvidia Jetson TX2.

I want to get the indices(row/column) of a set of points (with known X,Y,Z tuple values) from the point cloud.

Is there any fast search algorithm instead of checking each of the points in the point cloud data?

Thanks in advance!

Answer 1

What data structures are you using? This will depend greatly on how you're able to manipulate the data. It might help to look into some ways of partitioning the space so that you can break it into smaller search spaces.

There's a lot of literature on space partitioning:

• https://en.wikipedia.org/wiki/Binary_space_partitioning
• https://en.wikipedia.org/wiki/R-tree
• https://en.wikipedia.org/wiki/Hierarchical_clustering
• https://en.wikipedia.org/wiki/Octree
• https://en.wikipedia.org/wiki/K-d_tree
• https://en.wikipedia.org/wiki/Bounding_volume_hierarchy

Video game physics addresses the issue of collision detection a lot, so if you poke in that direction you'll find lots of ways you might lay this data out and search it efficiently. The general idea is that you want to partition both the space and the points as evenly as possible. Since the points aren't perfectly evenly distributed in the space, this takes some wizardry. Once you decide how you'll organize that data, it's usually something along the lines of a tree search to find points in a given region.

Happy reading!