Point Cloud Downsampling

Downsamples a point cloud or LaserScan by keeping only every N-th point. More...

Functions
LaserScan RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const LaserScan &cloud, int step)
	Downsamples a LaserScan.
pcl::PointCloud< pcl::PointXYZ >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointXYZ >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointXYZ.
pcl::PointCloud< pcl::PointXYZRGB >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointXYZRGB >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointXYZRGB.
pcl::PointCloud< pcl::PointXYZI >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointXYZI >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointXYZI.
pcl::PointCloud< pcl::PointNormal >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointNormal >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointNormal.
pcl::PointCloud< pcl::PointXYZRGBNormal >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointXYZRGBNormal >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointXYZRGBNormal.
pcl::PointCloud< pcl::PointXYZINormal >::Ptr RTABMAP_CORE_EXPORT	rtabmap::util3d::downsample (const pcl::PointCloud< pcl::PointXYZINormal >::Ptr &cloud, int step)
	Downsamples a point cloud of type pcl::PointXYZINormal.

Detailed Description

Downsamples a point cloud or LaserScan by keeping only every N-th point.

This function downsamples a point cloud based on a specified step size. The method adapts based on whether the cloud is organized (2D) or unorganized (1D), and whether it's structured like a depth image or LiDAR scan.

• For unorganized point clouds: Every step-th point is retained linearly.
• For organized point clouds with LiDAR-like layout (e.g., 2048x64): Downsampling is performed along the "long" dimension, preserving the ring structure.
• For depth-image-like organized clouds (e.g., 640x480): Downsampling is performed in both row and column directions, similar to image decimation. Step size must divide both width and height exactly.

Parameters

cloud	The input point cloud to downsample.
step	The decimation step. Must be greater than 0.

Returns

A new point cloud containing only the sampled points.

Note

If step <= 1 or the cloud has fewer points than step, the function returns a copy of the input cloud.

Exceptions

Assertion

failure if step <= 0 or, in the case of depth-image-style clouds, if the width and height are not divisible by step.