SEG-Y Data. Anyone tried it?

[Kieran Blacker]

Hi, 

Has anyone, to anyones knowledge, tried to load geophysical seismic reflection data stored as a SEG-Y file? Here's a brief summary of what a SEG-Y is. Its basically a header file consisting of XY geometry and properties, with a binary data trace then assigned to each header. 

Seems like hyperspy would be just the thing for dealing with large 3D datasets. 

Cheers, 

Kieran 

[Thomas Aarholt]

Hi Kieran,

I don't know of anyone who has written one specifically for hyperspy, but a quick Google search for "seg-y python" reveals several readers that would probably be quite easy to implement.

If you just want to try to see how hyperspy works with the data, you could import the data into python, and then call "s = hs.signals.Signal1D()" (or 2D) on the numpy array. You could then manually calibrate the names, units, offsets and scale.

- Tom

--
You received this message because you are subscribed to the Google Groups "hyperspy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hyperspy-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Kieran Blacker]

Alrighty then, I'll give that a shot and see how things go! Thanks 

My ultimate mini-project aim is to produce a pseudo-seismic cube from a lattice of 2D lines. The youtube video for hyperspy convinced me that this was probably one of the better ways to go about doing this. Reckon that is correct?

Kieran

[Francisco]

Kieran, it's hard to judge if HyperSpy is the right tool for the job without knowing much about your data and the sort of analysis that you would like to perform.

Francisco

[Kieran Blacker]

Ok here goes. It's a 2D lattice of seismic lines that I would like to resample into a pseudo 3D dataset. XY coordinates locate a signal trace sampled in the time domain. Pretty large dataset, ~50GB total, but I've already cropped 500mb subsets. 

The geometry of the input data is irregular, as wonky lines, and this would need generalising/gridding also. I've tested interpolation between lines in commercial software and as far as I can tell it will work fine. The idea is to get all that fiddly, unwieldy 2D data and streamline it into a hypercube that is easier to work with. 

In terms of my python experience, well its mostly been signal processing and stats thus far. This is something totally new, and quite scary. 

[Tomas Ostasevicius]

Kieran,

Here are the functions that you will likely end up using to do the interpolation.

If the z coordinate is uniform throughout the dataset, the simplest approach would be to create a XY coordinate map, and then do the interpolation layer by layer down the z axis. If not, you will have to do it in all 3 dimensions at the same time, chunking the total dataset however you see fit.

Tomas