Partitioning point clouds

Often times, when processing our point clouds, we might need to perform operations that require accessing a local region of a point cloud or manipulating the neighborhood of specific points. Since point clouds store data in a one-dimensional data structure, these kinds of operations are inherently complex. In order to solve this issue, PCL provides two spatial data structures, named the kd-tree and the octree, which can provide an alternative and more structured representation of any point cloud.

As the name suggests, an octree is basically a tree structure in which each node has eight children, and which can be used to partition the 3D space. In contrast, the kd-tree is a binary tree in which nodes represent k-dimensional ...

Get ROS Programming: Building Powerful Robots now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.