Classifcation Statistics processing time + visualization

[Benjamin Jakimow]

Problem:

The Classification statistics app allows to select normal multiband images without any class label information.

In case of “enmap_berlin.bsq” this leads to an “endless” GUI-blocking processing in the main thread. Furthermore, histogram labels can not be read as well (see screenshot)

Recommendation:

• process in a parallel thread
• hide raster layers without class label info from the QgsMapLayerComboBox (via setExceptedLayerList())

‌

‌

[Andreas Janz]

That’s exactly what I reported in one of the last VidCons and via Email :-). So let’s discuss this next time. I create a Trello card.

[Benjamin Jakimow]

If there is no classification scheme information at all, there is nothing to discuss about, right?

‌

[Andreas Janz]

Technically, every unique value is interpreted as a category (aka a class). So, even for a hyperspectral image, you can count all the unique values and create the plot.
We haven’t decided, if a raster is only a valid classification, if it as valid category names and colors defined.

[Andreas Janz]

So, now we decided at the last Zoom VidCon (BJ and AR), that only a raster with a Paletted renderer is a valid classification.

If he really wants to, the user can still use a hyperspectral raster band, but now has to setup a Paletted renderer with many many classes first.

[Andreas Janz]

resolves #470

→ <<cset b6aa81b861ef>>