| Re: Relative distances between vector elements and search for matches in other vector. [message #80010] |
Thu, 26 April 2012 06:48  |
Tito
Messages: 16
Registered: March 2012
|
Junior Member |
|
|
First of all I want to thank you for you input!
@ David: ratio = xx[0] - xx ... How it is possible to be sooo easy?!?
I corrected this and now looks much better and understandable.
@ Gianguido: a = where( abs(ratio-ratio2[jjj]) LT 0.05 ,count) - this works the same as mine but it is much shorter! thanks for that!
I tested: good=abs(diff) LT 0.05 and count=total(good,2),
And this gives me, indexes of 1 and 0, and I want the indexes of ratio2, like:
0,1,2,4,5,8,9 and i get :
print,good
0,19,38,....719
The loop is working however thats maybe because the first line of the catalog and the first line of the spectrum "this time" are the same. e.g.
ratio = xx[0] - xx
0,-0,17,-0,34.....
and
ratio2 = ddd[0] - ddd
0,-0,1703,-0,22,-0,3403.... so it is easy to identify those differences which pass the (a = where( abs(ratio-ratio2[jjj]) LT 0.05 ,count)) criteria.
in other hand if they are shifted e.g. the first line from catalog is not the first identified line in the spectrum I am having a problem...
My goal is to identify lines from different catalogs with those from my spectrum based on their relative differences(distances) this is necessary because usually the lines from the spectrum are shifted but they have the same ratios between each other ( they might differ slightly) with those from the catalog.
So first I am finding the atmosphere lines on the spectra and what is left is the stellar lines which I will search again in catalog.
This is important for me because
1st. it will be automatic! ,
2nd. the method will be more accurate then if I have to select the lines manually. Sometimes there is 2 or more lines very close to each other and selecting is pain.
Now i probably need something that will find the same space between the two vectors.
I read about histogram from David's website but my head is about to explode.
I am still investigating, how to use 'histogram'...
Lets assume that I identify the telluric lines and now I have only the stellar which I will search from catalog and this time I know that the first line of the spectrum and the catalog are not the same. So I did this:
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
selected_dep = [] ; create empty arrays because I want to append (I dont know the result number of array elements)
selected_lines = []
ratio_sel_vald = []
for z = 0L,n_elements(ratio_stlines)-1 do begin
for zz = 0L,n_elements(ratio_stlines)-1 do begin
aa = where( abs((ratio_stlines[zz] - ratio_stlines[z]) - ratio_vald) LT 0.0025, count)
if count gt 0d0 then begin
selected_dep = [selected_dep,z2[aa]]
selected_lines = [selected_lines, z1[aa]]
ratio_sel_vald = [ratio_sel_vald,ratio_vald[aa]]
endif
endfor
endfor
;;;I have to use 'Uniq' because froim this 2 loops I have many Duplicates.
selected_lines = selected_lines[uniq(selected_lines,sort(selected_lines))]
ratio_sel_vald = ratio_sel_vald[uniq(ratio_sel_vald,sort(abs(ratio_sel_vald)) )]
And this is working! if I want to find which lines from the catalog are in my spectra. BUT i want to give to "selected_lines" not the "selected_dep"(depths) which are from the catalog, I want to associate the fonded lines with the depths from the spectra. (- And this is my biggest problem actually I dont know how to do this!)
Later I have to construct synthetic template(spectrum) based on that data, that will have the same depths and profile like those from the spectra BUT different wavelength. Then I will cross-correlate both synthetic mask + spectra in order to find the shift.
Thank you again for your time reading this and trying to help me!
Trifon
|
|
|
|
| Re: Relative distances between vector elements and search for matches in other vector. [message #80019 is a reply to message #80010] |
Wed, 25 April 2012 08:03  |
cgguido
Messages: 195
Registered: August 2005
|
Senior Member |
|
|
Also, I would also get nx=n_elements(xx) once and use it when I need it it.
Finally, your min for loop, I suspect one histogram could do it. but I am not sure what you're trying to do...
If you keep your loop then you could try this:
a = where( abs(ratio-ratio2[jjj]) LT 0.05 ,count)
and also avoid appending to stel_dep adn stel_wave by greating them outside the loop and filling as needed. Then clip one at the end.
PS: histogram wise... I'd start with something approximately like:
nr=n_elements(ratio)
nr2=n_elements(ratio2)
bigratio=rebin(ratio,[nr,nr2],/sample)
bigratio2=transpose(rebin(ratio2,[nr2,nr], /sample))
now you have two arrays that are the same size, so you can do:
diff=bigratio-bigratio2
;good is now filled with 1s and 0s
good=abs(diff) LT 0.05
;you can count them by using DIMENSION with total:
count=total(good,2)
...
On Wednesday, April 25, 2012 9:03:00 AM UTC-5, Trifon Trifonov wrote:
> Hallo all
> I am new in IDL but I am increasing my level everyday and I am so happy that I discover IDL!!!
> I stuck very badly in one algorithm that I am developing for my Pipeline.
> I am searching for a routine or method to:
>
> 1st. define the relative distances between vector elements ( e.g. distances between peaks in the spectrum(allready identified))
>
> 2nd. to identify the lines from another vector (bigger or smaller in array size)
> based on the criteria of the above (1st.)
>
> E.g.
>
> I have:
> RATIO_STLINES FLOAT = Array[21]
> 0.00000 -0.171038 -0.594733 -1.21922 -1.40923 -1.52994 -1.56168 -1.92495 -2.28294 -2.34803 -2.38444
> -3.15909 -3.46812 -3.61399 -3.97221 -4.09896 -4.27114 -4.43637 -5.69552 -6.05595 -6.23174
>
> RATIO_CAT FLOAT = Array[44]
> 0.00000 -0.209757 -0.249368 -0.930353 -1.05059 -1.21623 -1.21693 -1.74758 -1.75778 -1.77888 -1.78749
> -2.03075 -2.22571 -2.99081 -3.29190 -3.31060 -3.42393 -3.43413 -3.43844 -3.44364 -3.59758 -3.61648
> -3.76892 -3.80183 -3.92147 -4.26676 -4.53263 -5.02637 -5.31455 -5.35194 -5.52650 -5.59132 -5.75987
> -5.81128 -6.60440 -6.68362 -6.71653 -6.73373 -6.81295 -6.95529 -7.23017 -7.48834 -7.66608 -7.90535
>
> And the RATIO_CAT consist the lines from RATIO_STLINES (it is possible not all of them to be in the RATIO_CAT)
>
> So RATIO_STLINES is lets say shifted respect to RATIO_CAT and I want to identify lines based on their line ratios.
>
> I want to mention that I already made a working algorithm but I am 1000 % sure that the more experienced people here can help me to make it much more elegant and efficient. Later on I have a other problem on that I stuck but lets see can you help me on that and it is possible to solve it my self... if not I will post again :(
>
>
> Here is the code:
>
> ;Finding the peak of each line on the spectrum:
>
> peak = peaks(1-chip,0.9) ; finding the peaks from JJohnson IDL.
> ddd = xaxis[peak] ;wavelengths of the finded lines.
> lines = chip[peak] ;depts of the finded lines.
> ;xx - HITRAN lines
> ;z1- VALD catalog wavelengths
> ;z2 - VALD catalog line depths
>
> ;Finding the relative distance between the HITRAN lines (nm):
> ratio=fltarr(n_elements(xx))
>
> for j=0L, n_elements(xx)-1 do begin
> ratio[j] = xx[0] - xx[j]
> endfor
>
> ;Finding the relative distance between each line on the spectrum:
> ratio2=fltarr(n_elements(ddd))
>
> for jj=0L, n_elements(ddd)-1 do begin
> ratio2[jj] = ddd[0] - ddd[jj]
> endfor
>
> ;Finding the relative distance between the lines from the VALD catalog:
> ratio_vald=fltarr(n_elements(z1))
>
> for jjj=0L, n_elements(z1)-1 do begin
> ratio_vald[jjj] = z1[0] - z1[jjj]
> endfor
>
>
> ; Identify the telluric lines on the spectra. Only stellar lines will be left.
> telluric_depts = fltarr(n_elements(xx))
> telluric_waves = fltarr(n_elements(xx))
> stel_dep = []
> stel_wave = []
> stel_ra = []
> for jjj = 0L, n_elements(ratio2)-1 do begin
> a = where((ratio GT (ratio2[jjj] - 0.05)) and (ratio LT (ratio2[jjj] + 0.05)),count)
>
> if count gt 0d0 then begin
> telluric_depts[a] = lines[jjj]
> telluric_waves[a] = ddd[jjj]
> endif
>
> if count le 0d0 then begin
> stel_dep =[stel_dep, lines[jjj]]
> stel_wave =[stel_wave, ddd[jjj]]
> ;stel_ra = [stel_ra,ratio2[jjj]]
> endif
>
> endfor
>
> ;Finding the relative distance between the lines that left from the science spectrum:
>
> ratio_stlines = fltarr(n_elements(stel_wave))
> for k = 0L, n_elements(stel_wave) - 1 do begin
> ratio_stlines[k] = stel_wave[0] - stel_wave[k]
> endfor
>
>
> And I will stop here because I have more problems below...
>
> Thank you very much in advance!
|
|
|
|
|
|
|
|
| Re: Relative distances between vector elements and search for matches in other vector. [message #80144 is a reply to message #80010] |
Thu, 26 April 2012 14:30 |
cgguido
Messages: 195
Registered: August 2005
|
Senior Member |
|
|
Trifon, you're most welcome.
If you can estimate the max possible size of your result (even if you overestimate it by a lot) I would do:
selected_dep=fltarr(1e4)*0-1
;this creates a huge array with "impossible values" -1.
Then you fill it as you go with real values, And at the end just do:
selected_dep = selected_dep[ where(selected_dep NE -1)]
This is usually much faster!
In general, I am still confused about what you are trying to do. Could you give a simple example of input and output? Something like:
;inputs:
r1=[1,3,3,4,5]
r2=[2,-1,4,3]
;outputs:
3,4
Don't bother with technical (geological) terms, they just confuse non-specialists. Just give us the simple examples and we'll see if we can speed this up for you a bit.
Gianguido
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
