User-Defined Point Fields

In a LAS file, the header specifies a data record length, which is the number of bytes used per point record in the file. Note that this number could be larger than the minimal number of bytes needed to store the point format, and if this is the case, the point records will have "extra bytes" associated with them. The LAS 1.4 spec introduces an Extra Bytes VLR, which documents extra fields (which we will refer to as "user fields") associated to each point record with their name and data type, amongst other information. Note that in legacy versions, you could also store these "extra bytes" for point records, but there is no way to infer what type and what field names they have, so they are saved in the record as "Undocumented Bytes".

You can save user fields in your point cloud quite easily by including them as columns in your point cloud table. For example, you can save the user field "thing" along with your points and their classifications as follows:

pc = Table(position = rand(SVector{3, Float64}, 10), classification = rand(UIn8, 10), thing = rand(10))
save_las("my_pc.las", pc)

Under the hood, LASDatasets.jl will automatically create the appropriate header fields for data record lengths etc. and will also create the appropriate Extra Bytes VLRs and save them in your LAS file.

You can also add user fields to your LAS datasets as you go by calling add_column! with your new data and column name:

add_column!(las, :thing, rand(10))

This will add the user field "thing" to your dataset las and append the appropriate Extra Bytes VLRs to it.

Reading user fields from LAS files is just as easy, since you can simply specify the desired user fields in your requested columns, e.g.

las = load_las("my_pc.las", [:position, :classification, :thing])

Note that user fields can also be vectors of static vectors, and once again LASDatasets.jl will automatically save the appropriate Extra Bytes VLRs for you so you can specify them as you would any other user field:

using StaticArrays
pc = Table(position = rand(SVector{3, Float64}, 10), classification = rand(UIn8, 10), thing = rand(SVector{3, Float64}, 10))
save_las("my_pc.las", pc)
las = load_las("my_pc.las", [:position, :classification, :thing])

It's also worth mentioning that you can save "undocumented bytes" to any point as well by labelling them with the column :undocumented_bytes. In practice, it is possible to have point records with user fields and undocumented bytes, just one or the other, or neither (just the LAS point itself). For more details see here.