Point cloud data to CF-NetCDF

[lhmarsden]

Question

I am trying to write a program to convert point cloud data to CF-NetCDF.

What is a point cloud?

Imagine you have a structure - this could be a man-made structure like a building or a natural structure like a valley. A point cloud is a set of data points that repsent the 3D shape of that structure.

Point clouds in CF-NetCDF

There are different ways to represent the coordinate system of a point cloud. For natural structures a global coordinate system is often used (latitude, longitude, elevation). I envisage that this type of data would be quite easy to store in a CF-NetCDF file, using one dimenion (a counter for each point) and variables for latitude, longitude, elevation.

But what about coordinate systems relative to some known coordinate - e.g. X, Y Z coordinates relative to the location of the scanner or some other user-defined location?

Of course we could encourage people to use global coordinate systems. However, sometimes the data are very high resolution (e.g. 1 mm) and scanned objects can be very small. I suspect it will be more efficient to use an X, Y, Z coordinate system relative to some reference point.

Question: What is the best practice for using X, Y, Z coordinate systems relative to a defined point in CF-NetCDF? Consider that the files could be large, and storing latitude and longitude for every point might not be desirable.

Ideally the direction of X or Y should be defined too.

[lhmarsden]

Here is what I think would be suitable when using a global coordinate system:

netcdf test {
dimensions:
	point = 20 ;
variables:
	float longitude(point) ;
		longitude:units = "degrees_east" ;
		longitude:long_name = "longitude" ;
		longitude:standard_name = "longitude" ;
	float latitude(point) ;
		latitude:units = "degrees_north" ;
		latitude:long_name = "latitude" ;
		latitude:standard_name = "latitude" ;
	float altitude(point) ;
		altitude:units = "meters" ;
		altitude:long_name = "geometric height above geoid" ;
		altitude:standard_name = "altitude" ;
		altitude:positive = "up" ;

[JonathanGregory]

Dear Luke @lhmarsden

Your example is certainly fine in CF. Your point dimension is an example of a discrete axis (sect 4.5) and could also be labelled as a point feature type (sect 9.1).

Best wishes

Jonathan

[lhmarsden]

Interesting, thanks @JonathanGregory, I had wrongly assumed that a point feature type could only include 1 point. I will use this!

[JonathanGregory]

Yes, it's fine to use the point feature type for a collection of points. However, apart from labelling it with featureType, there isn't any difference from what you did in your example, which is already fine in CF e.g. compare with example H.1. It makes perfect sense to have a discrete axis with several auxiliary coordinate variables. The discrete sampling geometries of chapter 9 are most useful with all the other types, which contain more than one element per feature, because of the extra machinery they provide for storing a collection of them with different sizes in a single data variable. For point features, one size (unity) fits all!

[JonathanGregory]

I've marked this with the frequently asked question label, although I don't remember it being asked before, because it might be helpful to copy it to the FAQ.