[GRASS-user] Derivation of Landsat 8 Tasseled Cap coefficients for certain scene in Grass

Kozlova_Maria1 · June 9, 2015, 9:42pm

Dear all,
Asking for help in sudden difficult situation:
I need to derive proper, 'personal', Tasseled Cap coefficients for some scenes of my study area.
I know, such procedure can be done by Matlab, but unfortunately, my colleague who could do this have now some troubles and cannot help me at the moment.

I need to do this work but can use only Grass. So is it possible to derive my own TC coefficients for each scene using Grass? If so, which modules should I use?
Is it possible to do image Tasseled cap transformation with these new coeficients?
Am I right, that Grass module i.tasscap only provide TCT with use of standard TC coefficients (in the case of OLI images derived by M.H.A Baig et al, 2014)?

Would be appreciate for any information

--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ruš
š \------------------------------------/

NikosAlexandris · June 10, 2015, 12:10am

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 00:42:21 +0300]:

Dear all,
Asking for help in sudden difficult situation:
I need to derive proper, 'personal', Tasseled Cap coefficients for some scenes of my study area.
I know, such procedure can be done by Matlab, but unfortunately, my colleague who could do this have now some troubles and cannot help me at the moment.

I need to do this work but can use only Grass. So is it possible to derive my own TC coefficients for each scene using Grass? If so, which modules should I use?

First, what is the plan to derive the custom coefficients?

Is it possible to do image Tasseled cap transformation with these new coeficients?

Sure, as far as I understand, it's a simple sum. Something similar like (stolen
from the i.tasscap script):

r.mapcalc "output = k1 * in1band + k2 * in2band + k3 * in3band + k4 * in4band + k5 * in5band + k6 * in6band + k0"

Am I right, that Grass module i.tasscap only provide TCT with use of
standard TC coefficients (in the case of OLI images derived by M.H.A
Baig et al, 2014)?

I quickly checked the numbers in the script and the paper by Baig 2014.
They are the same for *only* Brightness, Greenness, Wetness, TCT4. The
coefficients for TCT5, TCT6 are missing (compare the numbers in the
paper, page 428 with
<https://trac.osgeo.org/grass/browser/grass/branches/releasebranch_7_0/scripts/i.tasscap/i.tasscap.py#L86>\).

Are these required? They can be added quickly I think.

Would be appreciate for any information

Hope the above helps a bit, Nikos

mlennert · June 10, 2015, 7:36am

On 09/06/15 23:42, Kozlova Maria wrote:

Dear all,
Asking for help in sudden difficult situation:
I need to derive proper, 'personal', Tasseled Cap coefficients for some scenes of my study area.
I know, such procedure can be done by Matlab, but unfortunately, my colleague who could do this have now some troubles and cannot help me at the moment.

I need to do this work but can use only Grass. So is it possible to derive my own TC coefficients for each scene using Grass? If so, which modules should I use?
Is it possible to do image Tasseled cap transformation with these new coeficients?
Am I right, that Grass module i.tasscap only provide TCT with use of standard TC coefficients (in the case of OLI images derived by M.H.A Baig et al, 2014)?

Yes. BTW, the module is actually "just" a python script [1], so you easily modify the coefficients if you want to.

Tassled cap coefficients are derived via principal component analysis + possibly some rotation. You can actually get pretty close by just doing your own PCA using i.pca. As Nikos said: it all depends on what you want to do.

Moritz

[1] https://trac.osgeo.org/grass/browser/grass/trunk/scripts/i.tasscap/i.tasscap.py

Kozlova_Maria1 · June 10, 2015, 8:48am

Dear colleagues,

Thank you for response.
a little bit later I will type the full procedure for derivation of TC coefficients.
PCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
The main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
Though I have to describe it in more details later...

--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

10.06.2015, 10:38, "Moritz Lennert" <mlennert@club.worldonline.be>:

On 09/06/15 23:42, Kozlova Maria wrote:

Dear all,
Asking for help in sudden difficult situation:
I need to derive proper, 'personal', Tasseled Cap coefficients for some scenes of my study area.
I know, such procedure can be done by Matlab, but unfortunately, my colleague who could do this have now some troubles and cannot help me at the moment.

I need to do this work but can use only Grass. So is it possible to derive my own TC coefficients for each scene using Grass? If so, which modules should I use?
Is it possible to do image Tasseled cap transformation with these new coeficients?
Am I right, that Grass module i.tasscap only provide TCT with use of standard TC coefficients (in the case of OLI images derived by M.H.A Baig et al, 2014)?

Yes. BTW, the module is actually "just" a python script [1], so you
easily modify the coefficients if you want to.

Tassled cap coefficients are derived via principal component analysis +
possibly some rotation. You can actually get pretty close by just doing
your own PCA using i.pca. As Nikos said: it all depends on what you want
to do.

Moritz

[1]
https://trac.osgeo.org/grass/browser/grass/trunk/scripts/i.t…

Kozlova_Maria1 · June 10, 2015, 10:08am

Dear Nikos,
TCT5 may contain atmospheric effects and TCT6 mainly represent noise. Though as far as I know it is not well explored if this information is quite useful.
Maybe it may be valuable in some cases....
Anyway first 4 components represent most image information.
--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

10.06.2015, 03:05, "Nikos Alexandris" <nik@nikosalexandris.net>:

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 00:42:21 +0300]:

Dear all,
Asking for help in sudden difficult situation:
I need to derive proper, 'personal', Tasseled Cap coefficients for some scenes of my study area.
I know, such procedure can be done by Matlab, but unfortunately, my colleague who could do this have now some troubles and cannot help me at the moment.

I need to do this work but can use only Grass. So is it possible to derive my own TC coefficients for each scene using Grass? If so, which modules should I use?

First, what is the plan to derive the custom coefficients?

Is it possible to do image Tasseled cap transformation with these new coeficients?

Sure, as far as I understand, it's a simple sum. Something similar like (stolen
from the i.tasscap script):

r.mapcalc "output = k1 * in1band + k2 * in2band + k3 * in3band + k4 * in4band + k5 * in5band + k6 * in6band + k0"

Am I right, that Grass module i.tasscap only provide TCT with use of
standard TC coefficients (in the case of OLI images derived by M.H.A
Baig et al, 2014)?

I quickly checked the numbers in the script and the paper by Baig 2014.
They are the same for *only* Brightness, Greenness, Wetness, TCT4. The
coefficients for TCT5, TCT6 are missing (compare the numbers in the
paper, page 428 with
<https://trac.osgeo.org/grass/browser/grass/branches/releaseb…>\).

Are these required? They can be added quickly I think.

Would be appreciate for any information

Hope the above helps a bit, Nikos

NikosAlexandris · June 10, 2015, 11:18am

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 11:48:44 +0300]:

Dear colleagues,

Thank you for response.
a little bit later I will type the full procedure for derivation of TC coefficients.
PCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
The main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
Though I have to describe it in more details later...

This introduction <http://www.sjsu.edu/faculty/watkins/tassel.htm> looks
good for me

Nikos

NikosAlexandris · June 10, 2015, 11:19am

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 13:08:54 +0300]:

Dear Nikos,
TCT5 may contain atmospheric effects and TCT6 mainly represent noise. Though as far as I know it is not well explored if this information is quite useful.
Maybe it may be valuable in some cases....
Anyway first 4 components represent most image information.

Sounds really close to PCA. Thank you Maria, Nikos

Kozlova_Maria1 · June 10, 2015, 11:50am

Yes. Though this paper is based only on MSS and TM data. Till now there are much more interesting developments very useful in land cover monitoring

--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ruš
š \------------------------------------/

10.06.2015, 14:13, "Nikos Alexandris" <nik@nikosalexandris.net>:

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 11:48:44 +0300]:

šDear colleagues,

šThank you for response.
ša little bit later I will type the full procedure for derivation of TC coefficients.
šPCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
šThe main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
šThough I have to describe it in more details later...

This introduction <http://www.sjsu.edu/faculty/watkins/tassel.htm> looks
good for me

Nikos

mlennert · June 10, 2015, 11:56am

On 10/06/15 13:18, Nikos Alexandris wrote:

* Kozlova Maria <kclo@yandex.ru> [2015-06-10 11:48:44 +0300]:

Dear colleagues,

Thank you for response.
a little bit later I will type the full procedure for derivation of TC coefficients.
PCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
The main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
Though I have to describe it in more details later...

This introduction <http://www.sjsu.edu/faculty/watkins/tassel.htm> looks
good for me

I don't have access to the paper used for the OLI version, but from the abstract the authors do "just" use PCA + rotation to get a result that is as close as possible to a direct tasselled cap transformation using the original 1970s coefficients.

And when you run PCA on MSS images you often get quite close results to what Kauth and Thomas arrived at through their clustering approach.

Moritz

Kozlova_Maria1 · June 11, 2015, 8:43am

Well...While I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
And thank you again for your support!

https://disk.yandex.ru/client/disk/Документы

And thank you again for your support!
--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ruš
š \------------------------------------/

10.06.2015, 21:58, "Kozlova Maria" <kclo@yandex.ru>:

Well...
While I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
And thank you again for your support!

--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ru
š \------------------------------------/

10.06.2015, 14:58, "Moritz Lennert" <mlennert@club.worldonline.be>:

šOn 10/06/15 13:18, Nikos Alexandris wrote:

šš* Kozlova Maria <kclo@yandex.ru> [2015-06-10 11:48:44 +0300]:

ššDear colleagues,

ššThank you for response.
šša little bit later I will type the full procedure for derivation of TC coefficients.
ššPCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
ššThe main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
ššThough I have to describe it in more details later...

ššThis introduction <http://www.sjsu.edu/faculty/watkins/tassel.htm> looks
ššgood for me

šI don't have access to the paper used for the OLI version, but from the
šabstract the authors do "just" use PCA + rotation to get a result that
šis as close as possible to a direct tasselled cap transformation using
šthe original 1970s coefficients.

šAnd when you run PCA on MSS images you often get quite close results to
šwhat Kauth and Thomas arrived at through their clustering approach.

šMoritz

Kozlova_Maria · June 11, 2015, 12:44pm

Dear colleagues,

I attach the document containing description of one of algorithms for calculating TC coefficients.
Can it be done by Grass?
--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ruš
š \------------------------------------/

11.06.2015, 11:43, "Kozlova Maria" <kclo@yandex.ru>:

Well...While I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
And thank you again for your support!

https://disk.yandex.ru/client/disk/Документы

And thank you again for your support!
--
Best regards,

Maria V. Kozlova

š /------------------------------------\
š Information System Lab. (ISysLab)
š State Oceanographic Institute,
š Kropotkinsky per., 6,
š MOSCOW, 119034,
š RUSSIA
š ------------------------------------
š tel. š š+7 499 246 6448
š fax. š š+7 499 246 7288
š mailto: škclo@yandex.ru
š \------------------------------------/

10.06.2015, 21:58, "Kozlova Maria" <kclo@yandex.ru>:

šWell...
šWhile I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
šAnd thank you again for your support!

š--
šBest regards,

šMaria V. Kozlova

šš /------------------------------------\
šš Information System Lab. (ISysLab)
šš State Oceanographic Institute,
šš Kropotkinsky per., 6,
šš MOSCOW, 119034,
šš RUSSIA
šš ------------------------------------
šš tel. š š+7 499 246 6448
šš fax. š š+7 499 246 7288
šš mailto: škclo@yandex.ru
šš \------------------------------------/

š10.06.2015, 14:58, "Moritz Lennert" <mlennert@club.worldonline.be>:

ššOn 10/06/15 13:18, Nikos Alexandris wrote:

ššš* Kozlova Maria <kclo@yandex.ru> [2015-06-10 11:48:44 +0300]:

šššDear colleagues,

šššThank you for response.
ššša little bit later I will type the full procedure for derivation of TC coefficients.
šššPCA itself cannot be applied here as it is based only on statistical information of the image. TCT is physically based transformation and PCA is only the first step.
šššThe main problem I see here is the 'soil line' which must be determined (manually or automatically) and PCA axes rotation to align 1st PCA component in accordance to the soil line.
šššThough I have to describe it in more details later...

šššThis introduction <http://www.sjsu.edu/faculty/watkins/tassel.htm> looks
šššgood for me

ššI don't have access to the paper used for the OLI version, but from the
ššabstract the authors do "just" use PCA + rotation to get a result that
ššis as close as possible to a direct tasselled cap transformation using
ššthe original 1970s coefficients.

ššAnd when you run PCA on MSS images you often get quite close results to
ššwhat Kauth and Thomas arrived at through their clustering approach.

ššMoritz

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

(attachments)

TCT_algorithm.doc (416 KB)

Kozlova_Maria1 · July 4, 2015, 9:26pm

Dear colleagues,

Some time ago I asked if it is possible to derive custom TC coefficients by Grass and described the first algorithm for derivation of TC coefficients in attached file.

The steps needed to derive TC coefficients for multispectral image are not completely uniform.
Here is the description of second algorithm I'd like to follow for derivation of TC coefficients

The second recipe, based on TM tasseled cap transformation.
(Suggested by Huang et al., 2002, Lobster and Cohen, 2007 and Baig et al., 2014)
1.
Selection of ‘dummy target' (as it is called in a paper by Baig et al., 2014) by applying of TM TC-coefficients directly to OLI or ETM+ data Here, in Grass, I can choose sensor type TM for OLI data, so this step is seems to be possible to do in Grass by i.tasscap.
2.
PCA (it is clear for Grass and easily can be done in the images of interest by i.pca).
3.
PR rotation of PCA matrix seems to be the most problematic step. Generally it can be done by the finding of best alignment of pca matrix created in step2 with previously created (in step1) dummy target. Such a way: - more then 4 TC components should be included to i.tasscap - some kind of raster comparison and relative normalization must be done, I guess, but I don't know, what tool exactly should be used.

A little question directly to Tomas Brunclic:
Dear Tomas: what do you think about the application of your script on RRN (i.grid.correl.atcor....) in this case? I would be very appreciate for your answer.

4.
Get TC-coefficients by applying i.mapcalc to original Landsat data and resulting PR rotated data Such a way, TC coefficients arise from: PCA axes rotation coefficients PR of PCA matrix coefficients.

P.S.: TCT of Landsat MSS data can be done as I see only by Jackson's algorithm. So still hoping for any help with these two TCT algorithms.

--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

11.06.2015, 18:32, "Kozlova Maria" <kclo@ya.ru>:

Dear colleagues,

I attach the document containing description of one of algorithms for calculating TC coefficients.
Can it be done by Grass?

--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

11.06.2015, 11:43, "Kozlova Maria" <kclo@yandex.ru>:
Well...While I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
And thank you again for your support!

https://disk.yandex.ru/client/disk/Доку%…

And thank you again for your support!
--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

10.06.2015, 21:58, "Kozlova Maria" <kclo@yandex.ru>:
Well...
While I'm trying to type the algorithm step by step you may look through the following papers which can hopefully clarify the subject
And thank you again for your support!

--
Best regards,

Maria V. Kozlova
/\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\\
Information System Lab\. \(ISysLab\)
State Oceanographic Institute,
Kropotkinsky per\., 6,
MOSCOW, 119034,
RUSSIA
\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
tel\.    \+7 499 246 6448
fax\.    \+7 499 246 7288
mailto:  kclo@yandex\.ru
\\\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-/
10.06.2015, 14:58, "Moritz Lennert" <mlennert@club.worldonline.be>:
On 10/06/15 13:18, Nikos Alexandris wrote:
\* Kozlova Maria &lt;kclo@yandex\.ru&gt; \[2015\-06\-10 11:48:44 \+0300\]:
Dear colleagues,

Thank you for response\.
a little bit later I will type the full procedure for derivation of TC coefficients\.
PCA itself cannot be applied here as it is based only on statistical information of the image\. TCT is physically based transformation and PCA is only the first step\.
The main problem I see here is the &#39;soil line&#39; which must be determined \(manually or automatically\) and PCA axes rotation to align 1st PCA component in accordance to the soil line\.
Though I have to describe it in more details later\.\.\.
This introduction &lt;http://www.sjsu.edu/faculty/watkins/tassel.htm&gt; looks
good for me :\-\)
I don't have access to the paper used for the OLI version, but from the
abstract the authors do "just" use PCA + rotation to get a result that
is as close as possible to a direct tasselled cap transformation using
the original 1970s coefficients.

And when you run PCA on MSS images you often get quite close results to
what Kauth and Thomas arrived at through their clustering approach.

Moritz
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

Tomas_IV · July 8, 2015, 8:22am

Dear Maria,
I have found this note by chance, because I do not follow all the traffic in the GRASS mailing list.
I have to admit I have some difficulties to grasp the meaning of the steps you describe, for example the abbreviations PR and RRN, will have to read the articles you mention first. Generally, if the process includes radiometric normalization, and correlation-based normalization is applicable, use of my script should be possible.
But for now, I have found full text of the article Baig et al., 2014, on Researchgate: https://www.researchgate.net/publication/262005316_Derivation_of_a_tasselled_cap_transformation_based_on_Landsat_8_at-_satellite_reflectance
Hope it helps.

Tomas B.

···

Dne 4.7.2015 v 23:26 Kozlova Maria napsal(a):

Dear colleagues, 

Some time ago I asked if it is possible to derive custom TC coefficients by Grass and described the first algorithm for derivation of TC coefficients in attached file.  

 The steps needed to derive TC coefficients for multispectral image  are not completely uniform.
 Here is the description of second algorithm I'd like to follow for derivation of TC coefficients 
  
The second recipe, based on TM tasseled cap transformation.
  (Suggested by Huang et al., 2002, Lobster and Cohen, 2007 and Baig et al., 2014) 
1.
  Selection of ‘dummy target' (as it is called in a paper by Baig et al., 2014) by applying of TM TC-coefficients directly to OLI or ETM+ data   Here, in Grass, I can choose sensor type TM for OLI data, so this step is seems to be possible to do in Grass by i.tasscap.  
2.
 PCA (it is clear for Grass and easily can be done in the images of interest by i.pca). 
3.
  PR rotation of PCA matrix seems to be the most problematic step. Generally it can be done by the finding of best alignment of pca matrix created in step2 with previously created (in step1) dummy target. Such a way: - more then 4 TC components should be included to i.tasscap - some kind of raster comparison and relative normalization must be done, I guess, but I don't know, what tool exactly should be used.   

A little question directly to Tomas Brunclic:  
 Dear Tomas: what do you think about the application of your script on RRN (i.grid.correl.atcor....) in this case? I would be very appreciate for your answer.   

4. 
Get TC-coefficients by applying i.mapcalc to original Landsat data and resulting PR rotated data   Such a way, TC coefficients arise from: PCA axes rotation coefficients PR of PCA matrix coefficients.     

  P.S.: TCT of Landsat MSS data can be done as I see only by Jackson's algorithm. So still hoping for any help with these two TCT algorithms.

Kozlova_Maria1 · July 8, 2015, 8:37am

Dear Tomas,
Thank you for response!

PR is procrustes rotation
RRN correspond to Relative Radiometric Normalization
You may look through the articles stored here

https://disk.yandex.ru/client/disk/Документы

Muhammad Hasan Ali Baig did the procedure I describe with ENVI and mathlab. I'm trying to find if there is a way to do similar things with Grass

Thank for you

Maria

--
Best regards,

Maria V. Kozlova

/------------------------------------\
Information System Lab. (ISysLab)
State Oceanographic Institute,
Kropotkinsky per., 6,
MOSCOW, 119034,
RUSSIA
------------------------------------
tel. +7 499 246 6448
fax. +7 499 246 7288
mailto: kclo@yandex.ru
\------------------------------------/

08.07.2015, 11:22, "Tomáš Brunclík" <brunclik@atlas.cz>:

Dear Maria,
I have found this note by chance, because I do not follow all the traffic in the GRASS mailing list.
I have to admit I have some difficulties to grasp the meaning of the steps you describe, for example the abbreviations PR and RRN, will have to read the articles you mention first. Generally, if the process includes radiometric normalization, and correlation-based normalization is applicable, use of my script should be possible.
But for now, I have found full text of the article Baig et al., 2014, on Researchgate: https://www.researchgate.net/publication/262005316_Derivation_of_a_tasselled_cap_transformation_based_on_Landsat_8_at-_satellite_reflectance
Hope it helps.

Tomas B.

Dne 4.7.2015 v 23:26 Kozlova Maria napsal(a):

Dear colleagues, Some time ago I asked if it is possible to derive custom TC coefficients by Grass and described the first algorithm for derivation of TC coefficients in attached file. The steps needed to derive TC coefficients for multispectral image are not completely uniform. Here is the description of second algorithm I'd like to follow for derivation of TC coefficients The second recipe, based on TM tasseled cap transformation. (Suggested by Huang et al., 2002, Lobster and Cohen, 2007 and Baig et al., 2014) 1. Selection of ‘dummy target' (as it is called in a paper by Baig et al., 2014) by applying of TM TC-coefficients directly to OLI or ETM+ data Here, in Grass, I can choose sensor type TM for OLI data, so this step is seems to be possible to do in Grass by i.tasscap. 2. PCA (it is clear for Grass and easily can be done in the images of interest by i.pca). 3. PR rotation of PCA matrix seems to be the most problematic step. Generally it can be done by the finding of best alignment of pca matrix created in step2 with previously created (in step1) dummy target. Such a way: - more then 4 TC components should be included to i.tasscap - some kind of raster comparison and relative normalization must be done, I guess, but I don't know, what tool exactly should be used. A little question directly to Tomas Brunclic: Dear Tomas: what do you think about the application of your script on RRN (i.grid.correl.atcor....) in this case? I would be very appreciate for your answer. 4. Get TC-coefficients by applying i.mapcalc to original Landsat data and resulting PR rotated data Such a way, TC coefficients arise from: PCA axes rotation coefficients PR of PCA matrix coefficients. P.S.: TCT of Landsat MSS data can be done as I see only by Jackson's algorithm. So still hoping for any help with these two TCT algorithms.

,

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user