Point clouds: reading, clipping and thinning

1. 1Import (top bar, any mode) → Point cloud, or Point Cloud mode directly, and upload a LAS, LAZ, E57, PLY or XYZ/PTS file.
2. 2Name it, and state the coordinate system and vertical reference it is on.
3. 3Backsight reads the header and reports the point count, the extent, and what classifications the file carries.

What each format brings with it:

• E57, every scan in the file is combined, and each is moved by its own pose. A scan read without its pose sits where the instrument stood rather than where the site is. Returns the scanner flagged invalid are dropped and counted.
• PLY, ASCII or binary, either byte order. A classification is read when the file carries one; most exporters write none, and then every point is class 0 (never classified) and you can run the ground classifier yourself.
• XYZ and PTS, plain text, X Y Z on each line, separated by spaces, tabs or commas. A PTS point-count first line is recognised and skipped. Columns past the third are not imported.

• Colour by elevation, by classification, ground, vegetation and buildings told apart at a glance, or by the point's own colour when the file carries RGB, the way a drone cloud does. Both the colour and the density you settle on are saved with the cloud and used by every view of it: its own viewer, the Point Cloud tab, and the Draw tab's reference dots.
• Dot size, fine through heavy, is saved with the cloud too, so the same cloud reads the same way on the Draw canvas and in its own viewer.
• The viewer draws a decimated sample and says exactly how many of how many points it is showing. The Density control turns the sample up on demand, to half a million points, or every point when the whole cloud fits under that, re-read from the stored file each time.
• Densify this view re-samples inside exactly what is on screen, so density goes where you are looking. Zoomed in far enough that the view holds fewer points than the cap, you are seeing every single point there, and the label says so.
• Choosing a density reads the cloud once and keeps that sample as a file, so every later visit downloads it instead of re-reading the original, the same reason an orthophoto appears instantly the second time. Densify this view is situational and deliberately changes nothing that was saved.
• The classifications panel counts every class over the whole file, not just the sample.
• On the Draw canvas the cloud shows as reference dots, and the Point on cloud tool copies any dot into a real point, its exact coordinates and elevation, no surface needed first. Every sampled dot is an actual measured point from the file.

Clip and thin writes a new dataset and leaves the original alone.

1. 1Draw a boundary on the view, if you want to clip to an area.
2. 2Choose which classifications to keep, if you want only some.
3. 3Set a keep-every-nth if you want it thinner.
4. 4Derive. The new cloud records where it came from and what was done to it.