Georeferencing LAS files in geographic coordinates (lat/lon) or in cartesian (projected) ones?

Question

First of all, I am a newbie in GIS.

I have a lidar point cloud with GPS and telemetry data associated. Then, I transform coordinates of every point to an ECEF coordinate system and finally I transform the angular coordinates to a cartesian coordinate system, using the Local Tangent Plane (LTP) approximation. (This approximation uses the gps coordinates of first point as a reference system origin and then, reference the rest of points in respect with the first).

I am writing the LAS file using liblas, and I don't really know how to set the georeference information to the LAS file. I am writing the coordinates of every point in the LTP approximation so I have 2 questions:

1. Is it correct to write the points coordinates in respect to the LTP approximation or should I write coordinates in lat/lon (gps) units?. If this is correct, how can I set the georeferencing respect the LTP?.

2. If '1' is not correct and I must write the coordinates in lat/lon units, how can I set the georeferencing? Is writing an VLR with WGS84 value enough?

Answer 1

It depends on what version of the LAS specification you are using. If it is 1.3 or less, then the specs define georeferencing information using pre-defined (see specs) variable length records (VLRs) using the same format as the GeoTIFF:

Georeferencing for the LAS format will use the same robust mechanism that was developed for the GeoTIFF standard.

This format, though somewhat challenging to grok at times, is remarkably flexible. It relies on three defined tags called the GeoKeyDirectoryTag tag, which is like a table of contents for georef data, the GeoDoubleParamsTag tag, which is like a store of all double-precision values referred to in the GeoKeyDirectoryTag, and the GeoAsciiParamsTag tag, which similarly is used to store all ASCII (text) values. This site provides a good explanation and an example.

As of LAS v. 1.4 however, this method of storing georeferencing information was changed to favour the well-known text (WKT) format, also stored in defined VLRs, although the GeoTIFF format is still used for legacy:

The Coordinate Reference System (CRS) information for the point data is required for all data. The CRS information will be placed in Variable Length Records or Extended Variable Length Records (note that if the writer wishes to maintain legacy compatibility, then GeoTIFF in VLRs must be used). The CRS is represented by either GeoTIFF or Well Know Text as indicated by the WKT Global Encoding bit. Point Record Formats 0-5 can use either GeoTIFF or WKT (but not both simultaneously). Point Record Formats 6-10 must use WKT.

I see no reason given these flexible formats why you couldn't store point info in geographic coordinates (lat/long) but this would be fairly unusual for LAS data in that I've never seen it done previously. I imagine the reason is that LiDAR datasets tend to be of rather large scale (small spatial extent) and projected coordinate systems are therefore preferred. It makes calculating the distances between points, which is important for some algorithms (e.g. point classification or filtering), much easier.