Importing a point cloud into 3DM Analyst¶

3DM Analyst supports importing 3rd party point cloud formats in .las, .laz, and ASCII .xyz / .pts formats, optionally with intensity and colour information per point. To import the point cloud file, select the menu File | Point Cloud | Load.

After the import is complete, select the 3D View tab to view the imported point cloud data if you haven’t already.

To be useful in 3DM Analyst, the point cloud must be converted into a surface model. There are two methods to do this in 3DM Analyst:

Poisson surface reconstruction. This generates an implicit surface, and is best for point clouds imported from laser scanners because it is very good at filtering out the noise in the original point cloud. It sacrifices surface detail in exchange for a more accurate representation of the surface with low noise, albeit at a lower resolution that the original data.
Direct triangulation. This uses the point cloud points as-is, and is especially useful for checking the quality of the raw data as well as ensuring that the colour assigned to each point is exactly the same as the original file.

Poisson Surface reconstruction¶

Poisson Surface reconstruction is the preferred method for generating a surface model from point clouds, especially if the point cloud is noisy (e.g. LiDAR or Semi-global matching).

Poisson reconstruction is used if you select the Even Spacing option when you click Merge DTMs in the Multiple DTM Data Management dialog. It’s also used by DTM Generator if you use it to merge DTMs. For more details on the algorithm and what the various settings do, see the dedicated Poisson Surface Reconstruction article.

Direct Triangulation method¶

Direct triangulation is a 2.5-D operation that requires the points to be projected onto a plane to determine connectivity; it is used in the Multiple DTM Data Management dialog when you click Merge DTMs if the Even Spacing checkbox is not ticked:

3DM Analyst will automatically determine the best fit plane to use for triangulating the point cloud directly. Note that because this is a 2.5-D operation, the surface must not “overlap” itself when viewed in the direction normal to the projection plane. This means that it will work best if only relatively flat portions of the pit wall are used at a time for creating the DTM.

If the point cloud does wrap around (e.g. an entire pit, or an underground heading) then you can select which points to use for creating the DTM by using the Trim DTMs` icon in the toolbar (), left-clicking the mouse to create a bounding polyline, Ctrl + right click to define the final point in the polyline, and then choose Outside polygon to specify that you want to remove everything outside:

Then create the DTM using the Multiple DTM Data Management dialog, save the DTM, remove it, re-load the point cloud, and cut out another section to triangulate.

If you set Minimum Spacing to 0 then every point in the input cloud will be used for the output DTM. Since there is no colour information between points, this will maximise the resolution of the virtual texture, as nothing is removed.

Note that because this option takes every point in the point cloud as-is, you might find the DTM is rather noisy if you turn off the texture and look at the quality of the surface:

Noisy LiDAR data — Spikes in the raw LiDAR data.¶

The best way to deal with the spikes is to use select Build | Filter DTM Points… to bring up the DTM Filtering dialog and then under Spikes click the Remove button, then click OK and check the results. Repeat until no spikes remain.

Then, to reduce the overall noise, you can use the same dialog but this time click Smooth, click OK, and check the results. A few iterations will usually be enough to greatly improve the quality of the DTM.

Comparison¶

The image below shows a directly-triangulated point cloud from a Riegl Mini-VUX LiDAR system with Applanix APX-20 IMU PPK, flown at 84 m altitude. Points from two separate runs flown in opposite directions are included, making it quite easy to see the individual scan lines due to systematic height differences between them as a consequence of the accuracy limits. The average point spacing is about 140 mm and the noise in the vertical direction is about 20 mm. Moving the mouse over the image will show the result of applying 3DM Analyst’s smoothing algorithm three times; it clearly improves the ability to see the underlying surface:

Directly-triangulated raw LiDAR data.
Move mouse over image to see the effect of 3 × smoothing.

For comparison, below is the smoothed directly-triangulated surface model again, but this time moving the mouse over it will show the result of applying Poisson Surface Reconstruction to the same point cloud with a minimum point spacing of 150 mm and a point weighting of 0. (See Poisson Surface Reconstruction for an explanation of these values.)

Directly-triangulated and smoothed LiDAR data vs Poisson reconstruction.
Move mouse over image to see the Poisson surface.

The Poisson surface is slightly smoother, with the biggest difference in the vegetation running across the middle of the image where the points are the most “spiky” due to some points being on top of the vegetation and some on the ground below; in the other areas the results are pretty similar.