Downsampling LiDAR Data

LiDAR data is captured as a dense point cloud, where each point is one reading from the laser scanner. Depending on the area of the survey, it might be necessary to downsample the points in order to increase the efficiency of data processing. This article outlines the 2 tools offered in Micromine Origin for downsampling point cloud data: 1) Downsample Data and 2) Point Cloud Filter (Downsample). 

Downsample Data

The Downsample Data tool is located in Design | Utilities | Point | Downsample Points... and Implicit | Surface | Downsample Points (shared tool across both ribbons). 

After selecting your input file, you are required to enter a Mesh size in grid units (meters or feet depending on your working coordinate system). This parameter defines the new resolution of the downsampled data. An example of how this is calculated is shown below:

1. A grid of cubes at the chosen mesh size is created covering the extents of the data - a mesh size of 5 would create cubes measuring 5m³.
2. For each cube, the input point closest to the centroid is kept.

Point Cloud Filter (Downsample)

Point Cloud Filter (Downsample) tool is located in Survey | Point Cloud | Downsample. This tool requires the Survey Module activated on the licence.

This tool provides more advanced methods than the Downsample Data tool; these methods are:
1. Decimate

Downsampling by an integer factor is referred to as decimation.

• Keep every: Select this option to downsample the point cloud by keeping every <n>th point, where <n> is the number you enter.

• Reduce the number of points to: Point cloud data often comprises millions to billions of 3D points. Select this check box to reduce the number of points by expressing a filter value of <n> million points. This will speed up processing for very large datasets.

2. Nearest Neighbour

Select this option to use a Nearest Neighbour interpolation algorithm.

• Voxel Centre: select the points nearest to either the Voxel Centre. The process will calculate the midpoint of the voxel. The nearest neighbour to the mid-point is kept.
• Voxel Centroid: select the points nearest to the Voxel Centroid. The process will calculate the centroid of the voxel as the average point of all the points inside the voxel. The nearest neighbour to the calculated centroid is kept.
• Resolution: Enter a value that will be the size of the grid voxels used (the same process as mesh size in the Downsample Data tool).

3. Poisson Sample

Poisson sampling is a classic sampling method that defines the minimum distance between points in a point cloud.

• Radius: Define the radius of the Poisson disk. Any points that fall within the disc radius will be discarded.

Want to learn more?

Click here to login to our Learning Management System

Click here to request access