I've found two sources that appear to provide easy to read explanation between Radiance vs. Irradiance and remote sensing reflectance and water leaving radiance.
Starting with Radiance vs. Irradiance:
Irradiance is simple: exchange of energy (in the form of photons) across a given area of flat surface per time. Radiance is more complicated: exchange of ...
You may consider using the R package hsdar as a starting point to carry out transformation of spectral response data as well calculation of vegetation indices. You will find additional information about this package in this article published in the Journal of Statistical Software. While the hyperSpec package might also be of use.
This tutorial provides ...
It would help if you could provide more information on your case.
Are you struggling with a particular application?
How does feature scaling relate to the rest of your workflow?
Based on what information you provided, I'll attempt a general answer:
I suggest using the first option, i.e. scaling each band individually.
Some background information:
I assume ...
I have found the answer:
For a classification task, feature scaling should be done for each individual pixel, not for the individual band. That means, we need to compute mean for the individual pixel over all bands since each pixel in B dimension represents a specific object which we want to be classified. Therefore we need to scale the observation not ...
You should use only the red line, the FWHM, as is recomended in USGS Spectral viewer(Here):
To simulate how a sensor detects a surface feature in a given spectral band, the feature's spectrum is "convolved" with the sensor's RSR. This is based on the average of the wavelength extents of the full width at half maximum values.
The only method that you mentioned that meets the criteria of being an ensemble method and providing variable importance is random forests. I would direct you to the parameter selection methods available in the rf.modelSel function in the rfUtilites package.
If by "ensemble" you mean, multiple modeling approaches, please abandon this idea. This imposes ...
As others have mentioned, wavelengths aren't part of the standard data model so for data in ENVI format it is best to extract from the associated '.hdr' file separately, which is just a text file and often contains more information than GDAL reads.
I wrote a library to read ENVI header files in Python which is availble from https://github.com/pmlrsg/...
What you need to do is parse the 'fmmyyddtnnpnnrnnrdn_v_.spc' file associated with your dataset. Each line in that file corresponds to a band in your image.
First column in the file is the 'Wavelength Center Position' and the second column is 'Full Width at Half Maximum', meaning the bandwidth. It should be noted that there may be more than two columns, ...
Your filtered collection is named 'S2', but you erroneously call the unfiltered 'HyperionCol' using your print statement, which consists of 82152 elements (too much to display).
Unfortunately, no images match with your area of interest and daterange. You can check that using ImageCollection.size().
var S2 = HyperionCol
// Filter collection by time
It looks like your bands are not in the correct order to create the false color composite you're after. In your image showing your output for the NIR, Red, and Green composite it looks like the NIR band is being rendered as Green.
By default the image will load in QGIS with the first band rendered as Red, second as Green and third as Blue. If you created ...
finaly I find out solution. If you have similar problem, you can load imaage to memmap like this. Then you can work with ENVI file and for saving just delete variable.
src = envi.open(filename_hdr)
img = src.open_memmap(writable=True)