1

Topic: Problems with GCP

Hi again

I am using the old version of cosis-corr, it is running in an 64bits Windows version (not sure this is the problem), and I am finding the following problems:

a/ once the tie points have been choosen the second step, converting tie points to gcps reports that points are out of the dem.
However I can check the relations beetween the DEM and the subset of the Spot scene using Globalmapper and the scene is nicely inside the DEM..
I have changed the envi header setting the "Map Info" parameter to 1, instead of 1.5 for DEM, SPOT subset and for the shading of the DEM (it has been made using the dem and the illumination elevation and azimuth reported in the ANC file). It solved the problem but then I set the geographical corners of the three elements: shade, dem and spot to start in 1 and the problem arises again.

b/ Once I have been able to get the gcps file with correct altitudes, I went to the gpcs optimization, but this time show the following message:

File generated the : Tue Mar 12 16:09:24 2013
;GCPS optimization report:
;GCP / Easting misregistration (meter) / Northing Misregistration (meter)
;     1          NaN          NaN
;     2          NaN          NaN
;     3          NaN          NaN
;     4          NaN          NaN
;     5          NaN          NaN
;     6          NaN          NaN
;     7          NaN          NaN
;     8          NaN          NaN
;     9          NaN          NaN
;    10          NaN          NaN
;    11          NaN          NaN
;    12          NaN          NaN
;    13          NaN          NaN
;    14          NaN          NaN
;    15          NaN          NaN
;    16          NaN          NaN
;    17          NaN          NaN
;    18          NaN          NaN
;    19          NaN          NaN
;    20          NaN          NaN
;    21          NaN          NaN
;    22          NaN          NaN
;    23          NaN          NaN
;    24          NaN          NaN
;    25          NaN          NaN
;    26          NaN          NaN
;    27          NaN          NaN
;    28          NaN          NaN
;    29          NaN          NaN
;    30          NaN          NaN
;    31          NaN          NaN
;    32          NaN          NaN
;    33          NaN          NaN
;    34          NaN          NaN
;    35          NaN          NaN
;Easting mis-registration in meters (Average / Standard Deviation):           -NaN          -NaN
;Northing mis-registration in meters (Average / Standard Deviation):           -NaN          -NaN
;Norm mis-registration in meters (Average / Standard Deviation):      0.0000000     0.0000000
;
;Files:
;- Image: E:\Yesa\spotpeque2
;- Reference Image: E:\Yesa\bigmdtshade.tif
;- DEM: E:\Yesa\bigmdt05
;- Ancillary file: E:\Yesa\037-264_070906_0_T\S071015115553350\SCENE01\base.anc
;- GCPS/Tied Points/ICP file: E:\Yesa\peque2.pts
;
and this is the end.

I am trying to download the new version of COSIS-corr, I went through the registration procedure and I can see that an email will be sent to my email to download before next saturday March ,16. But I do not get any email at any of my three attempts.
My address is right now: rvgeigen@movistar.es, it may be that the mail have been sent to my old address.

I know that sounds like a "comedie d'errors" and present my excuses... but such is life!

Best regards
Carlos

2

Re: Problems with GCP

Did you select your tie points from a file that was a subset? If so, you should make sure that the xstart and ystart in the envi header file are zero, else you'll have a problem. You should never modify the map info.

Sebastien

3

Re: Problems with GCP

Yes, you are right, I set xstart and ystart to 1 for both raw scene, dem and shaded dem ! What about the downloading of the new COSIS-corr version? I do not receive the email for downloading, It is possible that the answer goes to the old address eigen@ya.com instead of the present address rvgeigen@movistar.es this is a problem because the area under treatment is composed by flat areas with step mountains....
Thanks for your advice
Carlos

4

Re: Problems with GCP

Well, yes and no... this time with the SPOT subset with 0 at  X and Y start everything works quite right and I have the depth from the mdt in the GCPS file. However the optimization for SPOT yields the same output as before. I am dealing with a DEM of 5m pixel size and a sub scene from an original SPOT cropped with ENVI . Both shaded dem and dem are projected as UTM zone 30 N with datum WGS-84. The SPOT image is "native" "pseudo geographic" datun WGS84 and pixel sizes: 0.00004659 degrees in X and 0.00004659 in Y.

The most interesting area is a dam beetween two ranges in which slopes heavy rock slides have been detected but the height of the water level  is correctly placed in the DEM.  I am taking 28 tie points avoiding the area of suspected displacement and RMS is 7.25m. In the GCPS optimization I choose the statistical correlator with a window size of  16, of course I have checked the frequency correlator and different window sizes and ranges from 15 upt to 30m with no success.

I have the same SPOT images in tif format and projected into UTM zone 30 WGS84, in theory these tifs have been orthorectified with same DEM that I am using, but I am reluctant to use them as the ANC file would be out of use... Is it correct or there is any way around?

A I say before I am using the 2009 version of COSIS-corr, no way to download the new one, which may be a trouble because most places at Spain are a mixture of flats and high sloping ranges.

Thanks again for your help and excuses for my repetitive enquiries..

Carlos

5

Re: Problems with GCP

I changed the dem file for a 25m pixel size, just in case that small pixel sizes are the trouble, but to no avail. I have tested almost any combination of X,Y start

SPOT XY --0
Shade XY- 0,1,1.5
mdt XY----0,1,1.5
In theory the correct one should be :
SPOT XY --0
Shade XY--1
mdt XY---- 1
But at any case the optimization pass fails off. The master image appears but not the slave one and there is no estimation of the X nor Y error.
I cannot think on anyother change.
Any idea??

Carlos