MATLAB hyperspectral toolbox issues

The MATLAB hyperspectral toolbox add-on to the Image Processing Toolbox seems pretty new and correspondingly buggy.

1) It appears incapable of reading in an image with an ENVI image with header, for example the command

>> hcube_mat = hypercube('AVIRIS_NG_refCube.img','AVIRIS_NG_refCube.hdr');

generates the following errors

Error using enviinfo (line 53)

File type number: char not supported.

Error in hypercube>getMetaforImg (line 1227)

header = enviinfo(imgFile);

Error in hypercube (line 304)

obj.Metadata = getMetaforImg(hdrfile, 'envi');

2) The Spectral Library capability seems a mess. I see no way of reading in a new spectral library other than the default ECOSTRESS library. Am I missing something here with regards to reading in my own ENVI spectral library?

Answer 1

[William]

Thanks for your help with this issue. I'm attaching a *.mat file with those 3 variables.

Answer 2

[Subhadeep Koley]

1) The error is originating from line number 53 of "enviinfo.m".

It seems that you have one local custom function in the name of "enviinfo.m", which is conflicting with the Hyperspectral Imaging Library's own enviinfo(__) function. There is no runnable code in line 53 of Hyperspectral Imaging Library's enviinfo(__) function. You can verify it by looking at the function defination of Hyperspectral Imaging Library's enviinfo(__) function.

Either rename/delete your local custom "enviinfo.m" function or remove it from MATLAB path, and then try again.

rmpath('thePathWhereCustomEnviinfoResides')

2) You can use the Hyperspectral Imaging Library's enviinfo(__) function to read metadata from your header file (.hdr file), and then use multibandread(_) to read the binary file containing your ENVI spectral library.

E.g.,

% Read the header file
info = enviinfo('enviSpectralLibrary.sli.hdr');

% Read the binary data file
lib = multibandread('enviSpectralLibrary.sli',[info.Height, info.Width, info.Bands],...
    info.DataType, info.HeaderOffset, info.Interleave, info.ByteOrder);

Hope this helps!

[William]

Thanks I did, in fact, have a third party enviinfo.m file and renaming it allowed the new hypercube function to work and likewise the steps for reading in the spectral library worked... THANKS!

[William]

Working with this a bit more, I see that the spectra read in from the ENVI spectral library via the suggested procedure are just read in as array. That's fine as far as it goes, but is there any mechanism for getting it into the structure format that your toolbox seems to want other than laboriously entering in 15 fields for each entry?

[Subhadeep Koley]

@William Most of the Field Names returned by readEcostressSig(_) are optional extra information. Only Wavelength and Reflectance fields are required to perform spectral matching.

Therefore, you can rearrange the custom ENVI spectral library as a 1-by-K structure array, where K is the number of spectrum present in your custom ENVI spectral library. Each element of the1-by-K structure array must contain at least two fields i.e., Wavelength and Reflectance. Then you can call spectralMatch with the syntax below,

scoreMap = spectralMatch(customEnviLibStructure, hCube); % Here hCube is a hypercube object

Consequently, entering in 15 fields for each entry is not necessary for performing spectral matching.

[William]

Thanks for the answer on that, but I'm still getting erroneous results. I build a spectral library with the two fields of Wavelength and Reflectance... putting those in for each entry in the library (and Wavelength is a one column vector that I replicate for each entry and each entry has its number of bands (for rows) and number of reflectance spectra vectors for each Reflectance entry. Then doing a match of a vector against that library with the command:

SAM_Values = spectralMatch(Fe2Fe3Mins2,sp202,wav,'Method','sam');

where Fe2Fe3Mins2 is the library structure, sp202 is a single column reflectance spectrum of the "unknown" and wav is the wavelength. I get a value for each entry, but they're one of two numbers 3.01 or 0.13 for each entry, so that's clearly not right. Any ideas?

[Subhadeep Koley]

@William, Is it possible for you to attach the Fe2Fe3Mins2, sp202, and wav variables as a .mat file here?

Without the actual data, it's difficult to comment on the issue.

[William]

I thought I responded to this, but maybe it didn't go out:

Thanks for your help with this issue. I'm attaching a *.mat file with those 3 variables.