Wavelength in X-axis of Spectra instead of band nb.

[Agustin Lobo]

I'm just trying EnMAP-Box and cannot find the way to set the X-axis of the spectral plots in Wavelength instead of band number. Is there a way? Thanks

[Benjamin Jakimow]

is part of the next release and already tested.

[Agustin Lobo]

and how is the wl of each band specified? can the box use an envi hdr file? or just a txt file with 2 columns, band and wl?

[Benjamin Jakimow]

Yes, it uses the ENVI header file specification to define wavelength https://www.harrisgeospatial.com/docs/ENVIHeaderFiles.html

Let us know if there are other common ways to define wavelength for raster images and spectral library formats, we can implement them as well.

[Agustin Lobo]

I'm perfectly happy with envi's hdr, it is a "de facto" standard for hyperspectral metadata. I was actually worried that you might enforce a better (e.g. xml) but new (nobody using) format. I understand that, for enmap-box, this hdr does not require the companion file to be in "raw" format. I mean, I can have a geotif image "myimage.tif" and a companion "myimage.hdr" to describe the hyperspectral information.

[Agustin Lobo]

Benjamin, In case we have e.g. a geotif multi or hyperspectral image (ima.tif), should we make a full hdr companion file (ima.hdr) to include wavelength description? In that case, what should we state in the "file type" field? Also, the fact that we have ima.tif and ima.hdr, is not going to confuse gdal that would try to open ima.tif as en envi file? Perhaps the best is if you could share an example of a non-envi image file and its hdr companion file.

[Agustin Lobo]

I've made geotif version of hi_res and its hdr companion file with (guessed) wavelength information
https://www.dropbox.com/s/hc5z8v9yhar7b7r/hires_berlin2.tif
https://www.dropbox.com/s/2at6z71eiqsxcf6/hires_berlin2.hdr

I open hires_berlin2.tif in enmap-box and the spectral plot has no wavelength in the x-axis.
Could you explain how to get wavelength info in the spectral plots for non-envi files?
Thanks

[Benjamin Jakimow]

In this case we have the problem that a *.tif, which has it's internal way to store metadata, is combind with an ENVI header. QGIS and EnMAP-Box API automatically opens it with the GDAL TIFF Driver and those don't care about ENVI header file infos.

Solution 1:

Write your metadata as GDAL auxiliary metadata into a hires_berlin2.tif.aux.xml.

<PAMDataset>
  <Metadata domain="ENVI">
    <MDI key="wavelength">{ 0.650, 0.550, 0.470000}</MDI>
    <MDI key="wavelength_units">Micrometers</MDI>
  </Metadata>
</PAMDataset>

The hires_berlin2.tif.aux.xml will be used to store other information as well, e.g. QGIS-based raster histogram statistics.

Solution 2:

Convert the hires_berlin2.tif into an ENVI-like hires_berlin2.bsq, which will let QGIS / EnMAP-Box / GDAL opened the binary file using the ENVI driver.

[Agustin Lobo]

Hmmm... If I understand correctly, Solution 2 is not a solution: it is a change of format (I've tried just changing the extension from tif to bsq and I do not get to trick qgis: the file is still recognised as tif). As hyperspectral images use to be big, changing the format to have wavelength info in enmapbox is not a solution. We must avoid having to duplicate potentially huge files.

Regarding Solution 1, it would be greatly appreciated a tool to write that info to the xml file from info provided by the user as a 2 columns txt file or from a built-in list of sensors (i.e. EnMap, ASTER, S2 etc). I find easy writing text files (even writing a simple function in R) but writing (and potentially damaging) xml files is not as easy.

In any case, Solution 1 is not really as simple as having the required info in a companion hdr file, that is a simple text file to which everyone is used to and that does not interact with other qgis functionalities. I would favor

Solution 3

let enmapbox search the wavelength info in an optional companion file named i.e. *.hwl that would keep the fields wavelength units, band names, wavelength and fwhm following the same convention as hdr files.

[Agustin Lobo]

I'm actually trying Solution 1 and do not get it to work with a hyperspectral tif image. Please see here the xml file and an screenshot of how I get the plot:
https://www.dropbox.com/s/cxuwldjns3a3vnk/REFL_FX17_DeceptionIsland24_2018-04-06_08-39-09.tif.aux.xml
https://www.dropbox.com/s/4a4rp2trhwl3cis/REFL_enmapbox.JPG

[Benjamin Jakimow]

Plot and header look fine, but without REFL_FX17_DeceptionIsland24_2018-04-06_08-39-09.tif I can not test it in my EnMAP-Box.

However, can you change the x-axis units via the plot context menu -> X Axis > Unit?

[Agustin Lobo]

Weird, I get it very different: https://www.dropbox.com/s/o4l64qdfp22heoy/Screenshot_2019-03-28_08-14-39.png

I've put here the image in case you can try:

xml: https://ictjabox.ictja.csic.es/index.php/s/xxYVopzjhwFFfAB

geotif: https://ictjabox.ictja.csic.es/index.php/s/1lmx94Xh5i4DCjc

Thanks

[Benjamin Jakimow]

The problem is that REFL_FX17_DeceptionIsland24_2018-04-06_08-39-09.tif.aux.xml defines 1024 wavelength values, but REFL_FX17_DeceptionIsland24_2018-04-06_08-39-09.tif consists of 224 bands only. This way the EnMAP-Box can not match the wavelength value to the existing raster bands.

I'll add a warning to the source code and you just need to ensure that the number of float specified in

<MDI key="wavelength">{935.61,939.06, ...}</MDI>

is same as the number of bands.

[Agustin Lobo]

I confirm it works with the correct nb. of wavelength values.
So it was my fault. I probably Ctrl-pasted several times the vector of values. I apologize.

Now that I have wavelength in the X-axis, I miss band numbers... Perhaps the little window in the plot could have x in wavelength and band nb. in (), ie:

x: 850 (35)
y: 0.345

Thanks

[Agustin Lobo]

I understand the original issue has been addressed and this ticket can be closed

[Benjamin Jakimow]

I opened issue #267 to address the band number issue