findimageoverlaps errors when using HiRISE stereo pairs (SOLVED)

[rhemmi]

Dear Isis users,

I'm an ASP (Ames Stereo Pipeline) user and going to perform bundle adjustment using nonprojected HiRISE images (created by hiedr2mosaic.py command).
However, when I run findimageoverlaps command, I receive some error messages.

$ ls *.mos_hijitreged.norm.cub > cube.lis
$ findimageoverlaps fromlist=cube.lis overlaplist=overlap.lis
**USER ERROR** File [PSP_XXXXXX_XXXX_RED.mos_hijitreged.norm.cub] can not be added to serial number list.
**USER ERROR** Invalid serial number [Unknown] from file [PSP_XXXXXX_XXXX_RED.mos_hijitreged.norm.cub].

This is the print.prt output shown below.

 Group = UserParameters
    FROMLIST    = cube.lis
    OVERLAPLIST = overlap.lis
    DETAILED    = false
  End_Group

  Group = Error
    Program = findimageoverlaps
    Class   = "USER ERROR"
    Code    = 2
    Message = "File [PSP_XXXXXX_XXXX_RED.mos_hijitreged.norm.cub] can not be
               added to serial number list"
    File    = SerialNumberList.cpp
    Line    = 169
  End_Group

  Group = Error
    Program = findimageoverlaps
    Class   = "USER ERROR"
    Code    = 2
    Message = "Invalid serial number [Unknown] from file
               [PSP_XXXXXX_XXXX_RED.mos_hijitreged.norm.cub]"
    File    = SerialNumberList.cpp
    Line    = 149
  End_Group
End_Object

Do you have any idea to solve this problem?

Best regards,
Ryodo

[JohnVV]

the mailing list would be the best place to get help
( that is not to say here is not -- some of us DO use it )
but the EXPERTS are on the mailing list
http://ti.arc.nasa.gov/tech/asr/intelligent-robotics/ngt/stereo/#Mailing

[arkantos84]

Hi

I write here because I have the same problem of rhemmi. Reading the mailing list of ASP, it was noticed that this problem is caused by a bug of ISIS. Unlike other images (CTX, MOC etc), HiRISE metadata doesn't contain serial number. Hiedr2mosaic.py is a python command that use ISIS procedures for build a HiRISE non-projected mosaic starting from EDR files...

Suggestions for alternative methods?

Thanks in advance

Alessio

[rhemmi]

John,

To tell the truth, I'm subscribing to the ASP mailing list you recommended.
And I decided to make this post when I saw the exchange of emails below between Alessio and Ross (one of the ASP EXPERTS).

> When I try to launch findimageoverlaps from=cube.lis overlaplist=overlap.lis an error occurred (The xxx.cub serial number (unknow) does not exist) and i cannot go forward, because overlap.lis is not created.

Hmm, that's an ISIS problem, then.  The only thing I can say is that it *should* work.  If it isn't, then you can take that problem to the ISIS forums.

Supplementary information:

The hiedr2mosaic.py command contains the eight ISIS applications (hi2isis, hical, histitch, spiceinit, spicefit, noproj, hijitreg, and handmos).

And I know the other two processes to make non-map projected HiRISE mosaic.
1. HiRISE Tutorial
http://isis.astrogeology.usgs.gov/IsisWorkshop/index.php/Working_with_Mars_Reconnaissance_Orbiter_HiRISE_Data
2. hi4socet.pl (perl scripts for SOCET SET)
http://webgis.wr.usgs.gov/pigwad/tutorials/socetset/SocetSet4HiRISE.htm

I've not tried findimageoverlaps yet using the outputs created from these two processes.

[tbecker]

Going by the information supplied on the processing involved with the HiRISE images, (hi2isis, hical, histitch, spiceinit, spicefit, noproj, hijitreg, and handmos), I know for certain that findimageoverlaps will not work following 'handmos'.  The serial number is constructed from keywords in the Instrument Group on the input image labels...when any ISIS3 image is mosaicked using 'handmos', 'automos', and 'mapmos'...the Instrument Group is removed from the output labels, which is why you are seeing the 'invalid serial number' message.

Internally, we have had success with findimageoverlaps with HiRISE data following 'histitch, spiceinit and spicefit'.  If even possible, I don't think that we have established a procedure for images post noproj and hijitreg.

[rhemmi]

Becker,

Thank you very much indeed.
Alright... I'll try to make control points without using findimageoverlaps, autoseed, and pointreg. Probably only with qnet and/or qtie manually.

If you know how to inject a bogus serial number into the label of the mosaicked HiRISE image to perform findimageoverlaps properly, please let me know.

Ryodo

[tbecker]

Unfortunately you will encounter the same problem with qnet and qtie.  The entire 'control process' through jigsaw relies on the serial number.  The infrastructure established for control points and jigsaw requires image data to be in a 'native' instrument form (in the case of HiRISE, the histitch output with spice).  I will network with internal staff that might know if it is possible with noprojed-hijitreg output (meaning the files that are input to handmos); I suspect that it will not be possible.

[rhemmi]

Thank you for your quick response, Becker.

I see.  I don't perform a bundle adjustment on HiRISE mosaics for the present.

Ryodo

[Zack Moratto]

I took my own stab at this problem and documented it on my blog. Hopefully you guys don't mind the blog spam.

http://lunokhod.org/?p=820

In short, I wrote a new version of HiEDR2Mosaic that allows you to do the following:

Code: [Select]

Download all IMGs
hiedr2mosaic --stop LEFT*IMG
hiedr2mosaic --stop RIGHT*IMG
Do your control networking
Do your jigsawing
hiedr2mosaic --resume LEFT*cub
hiedr2mosaic --rseume RIGHT*cub
stereo left.cub right.cub HiRISE/output
point2dem HiRISE/output-PC.tif
At this point you've lost 12 hours of your life. Also, don't do the whole 'No-GCP' idea I wrote in the blog. That's a pet idea I'm still playing with. Instead just pick GCPs like a normal person who cares about their DEM results.

EDIT: You can download my new version of HiEDR2Mosaic at https://raw.github.com/NeoGeographyToolkit/StereoPipeline/master/src/asp/Tools/hiedr2mosaic.py.in .

[rhemmi]

Hello Zack,

Thank you very much for your help.
I've seen your message from the ASP mailing list and your blog today.

I followed your former articles about CTX bundle adjustment and I've created bundle-adjusted CTX DEMs successfully using control points, GCPs from THEMIS basemap, and cam2map4stereo.py.

Alessio,
I think cam2map4stereo.py works well if you picked control points and GCPs 'correctly'.
I mean you should be careful when you follow Zack's article. He omitted minor details.
You need to click "Save Measure" after every picking GCPs and to click "Save Measure" and "Save point" after every changing PointType.

With best regards,
Ryodo

[rhemmi]

Hi Zack,

When I run autoseed, I get the error message below.

**USER ERROR** The ouput control network is empty. This is likely due to the input cubes failing to overlap.

Where is my mistake?
Before running autoseed, I run hiedr2mosaic, footprintinit and findimageoverlaps as usual.
The content of cube.lis is the only two lines of images which must share an overlap area (e.g. left_RED0 and right_RED1). I always make sure that in qview.

With best regards,
Ryodo

[Zack Moratto]

Your error is likely in your settings of Autoseed (usually the Autoseed.def file). Make sure MinimumThickness is a very small number (like .00001) because the HiRISE CCDs are very rectangular. Thickness is the ratio of width/height. You'll also want XSpacing to be a small number so that CCDs actually get points. I used 200 meters for my XSpacing and then 3000 meters for YSpacing.

If this doesn't fix your problem. Please post your Autoseed.def file, along with caminfo output for all of your CCDs.

Code: [Select]

parallel caminfo from={} to={.}_caminfo.log ::: *.cub

[rhemmi]

Thank you very much indeed, Zack.
You're right. MinimumThickness was 0.01, so I changed it and then I've succeeded in performing autoseed.

With kind regards,
Ryodo

[tbecker]

Thank you, Zack, for helping.

Tammy