Re: [OSSIM] Problem in geo-referecning from SOPT 5 Sensor Model

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

Hello All:

Try an update.  In your new model you will need to initialize the  
seed function in the updateModel call.  Look in ossimSpot5Model's  
updateModel call to see how to do this.  This appears to be faster  
for me.

take care

Garrett

On Dec 6, 2007, at 7:24 AM, Debajyoti Dhar wrote:

> Dear Garrett , Dave and all OSSIM Friends
>
> I tried Spot5 sensor model with commenting the lines as Garrett  
> mentioned, but
> it still has problem. What i could find that extrapolate() method of
> Spot5Model returing nan, which is the point that image bounds are  
> chopped.
> Another observation, worldToLineSample() in ossimSensorModel  
> doesnot converge
> even after 20 iterations ( most of the time .) Can we provide  
> nearby points to
> start with instead of midpoint of the image space in this method ??  
> This might
> solve problem of non-convergence. IRS-P4 OCM ( Ocean Color  
> Monitor ) sensor is
> simpler sensor with ground resolution of 360x240m, onlything i do  
> extra over
> Spot5 sensor model converting attitude values from ECI to ECF by using
> Sidereal Angle whic i compute inside the sensor model file. If U  
> want, i will
> send U metadata parser ( which is now simple ascii text ), and all  
> other
> related files for U to have a look. The problem of clipping of  
> sides are
> visivle for SPot5 as well as OCM.
> regards
>
> Debajyoti
>
> dav...@l-... said:
>
>> Hi Garrett,
>>
>> Hey just fyi:
>>
>> The current ossimSpot5Model overrides "extrapolate" to return  
>> nan...  So I'm
> sure that's why points outside the image bounds are clipped.
>>
>> I think that was done (by you maybe?) to fix the spot model from  
>> sitting
> there spinning.
>>
>> Maybe for that model we should override the worldToLineSample and  
>> up front
> make a very coarse (coarse grid)...
>>
>> Just a thought...
>>
>> Dave
>>
>>
>>
>> -----Original Message-----
>> From: Garrett Potts [mailto:po...@cf...]
>> Sent: Wed 12/5/2007 10:30 AM
>> To: de...@sa...
>> Cc: Burken, David @ LINKABIT; oss...@li...
>> Subject: Re: [OSSIM] Problem in  geo-referecning from SOPT 5  
>> Sensor Model
>>
>> Hello:
>>
>> I am thinking also that the problem is within the extrapolate call.
>> For fun,  In ossimSensorModel, try to comment out the extrapolate in
>> the worldToLineSample call.  This might be the problem for the edges
>> that you are seeing.  Look at :
>>
>>
>> ossimSensorModel::worldToLineSample( .... )
>>
>> Look toward the top of the function where you see:
>>
>> if( .... )
>> {
>> if(!(theBoundGndPolygon.pointWithin( ... )
>> {
>> }
>> }
>>
>>
>> Comment that out and see what you get
>>
>> Take care
>>
>> Garrett
>>
>> On Dec 4, 2007, at 11:09 PM, Debajyoti Dhar wrote:
>>
>>> Dear Dave and OSSIM friends ..
>>>
>>> Sorry for the delay in replying you about your suggestion as i was
>>> out of
>>> station for few days. I have tried with ossimCoarseGridModel as
>>> intermediate
>>> transformation i.e. from ossimSpot4Model to ossimCoarseGridModel to
>>> ViewProjectionTransformation , and it works with no cut on either
>>> side. I
>>> could get the correct geo-referenced image for the OCEANSAT Sensor
>>> model using
>>> the same technique.
>>> I could go thru' the details of  ossimSensorModel.cpp ,
>>> *ViewProjectionTranformation and also the *ImageRenderer Code , i
>>> feel that
>>> the seed point used is always the MidPoint of the
>>> ImageClipRectangle for
>>> computing View-To-LatLong-To-ImagePoint by iterative method. When
>>> we use
>>> midpoint as first point for iterative process, it does not converge
>>> even with
>>> 20 iterations , and  in this process, sometimes some valid  
>>> ImagePoint
>>> corresponding to valid GroundPoint(Lat-Long) becoming nan - thus
>>> making whole
>>>  tile as empty. Is my observation correct ?? If by any means, we
>>> can give one
>>> of the corner of a valid tile as initial point, then we can
>>> converge within 3
>>> to 4 iteration , and we not miss any partial tiles. Can U please
>>> think of any
>>> such solution ?? we probably have to modify a few files. In the
>>> mean time, i
>>> will try to embed ossimCoarseGridModel inside my OCENASAT
>>> SensorModel code, so
>>> that i can do parameter adjustment ( like roll-offset, pitch-
>>> offset, attitude
>>> rate etc. ) directly from ImageLinker which is now not happening
>>> correctly. I
>>> need you suggestion to overcome this issue.
>>> regards
>>>
>>> Debajyoti
>>>
>>>
>>> dav...@l-... said:
>>>
>>>> Debajyoti,
>>>>
>>>> I'm not sure if this is your issue but there is a known bug/
>>>> feature with
>>>> the image view projection transform using the corner points only
>>>> to set
>>>> the clip rect.
>>>>
>>>> It should be:
>>>>
>>>> if (imageProjection->isAffectedByElevation()) walkTheEdgesToGetClip
>>>> ()...
>>>>
>>>>
>>>> See:
>>>> http://trac.osgeo.org/ossim/ticket/1263
>>>>
>>>> I haven't implemented this because hot on the list is reworking the
>>>> image/projection handling.  Garrett and I have talked about but it
>>>> will
>>>> be a pretty major change to the core so it hasn't happened yet.
>>>>
>>>> Dave
>>>>
>>>>
>>>>> -----Original Message-----
>>>>> From: oss...@li... [mailto:ossim-
>>>>> dev...@li...] On Behalf Of Debajyoti  
>>>>> Dhar
>>>>> Sent: Monday, November 26, 2007 1:01 AM
>>>>> To: oss...@li...
>>>>> Subject: [OSSIM] Problem in geo-referecning from SOPT 5 Sensor  
>>>>> Model
>>>>>
>>>>>
>>>>> Dear Garrett, Dave and all OSSIM friends ..
>>>>>
>>>>> I have been trying to implement sensor model classes for older  
>>>>> IRS (
>>>>> Indian
>>>>> Remote Sensing ) Satellites , which are having ground resolution
>>>>> from
>>>> 56m
>>>>> to
>>>>> 360m. I am observing that after geo-referecing, the geo-referenced
>>>> image
>>>>> is
>>>>> getting cut from both ends - left and right by  apprx'ly 100 to  
>>>>> 300
>>>>> pixels. I
>>>>> went back and processed one SPOT 5 Image from Dimap metadata and
>>>>> corresponding
>>>>> image data. I found 50+ pixels are missing  from leftside of the
>>>>> geo-referenced output with respect to the input image, whereas  
>>>>> right
>>>> side
>>>>> is
>>>>> having no cut in pixels. IS there anything i  am missing ?? I am
>>>>> using
>>>>> OSSIM
>>>>> 1.6.6 code in linux.
>>>>> Can U please check and  write on what could have gone wrong ??
>>>>> regards
>>>>>
>>>>>
>>>>> Debajyoti
>>>>>
>>>>>
>>>> ------------------------------------------------------------------- 
>>>> --
>>>> ---
>>>> -
>>>>> This SF.net email is sponsored by: Microsoft
>>>>> Defy all challenges. Microsoft(R) Visual Studio 2005.
>>>>> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
>>>>> _______________________________________________
>>>>> www.ossim.org
>>>>> Ossim-developer mailing list
>>>>> Oss...@li...
>>>>> https://lists.sourceforge.net/lists/listinfo/ossim-developer
>>>>
>>>
>>>
>>>
>>> -- 
>>>
>>>
>>>
>>>
>>> -------------------------------------------------------------------- 
>>> --
>>> ---
>>> SF.Net email is sponsored by: The Future of Linux Business White  
>>> Paper
>>> from Novell.  From the desktop to the data center, Linux is going
>>> mainstream.  Let it simplify your IT future.
>>> http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4
>>> _______________________________________________
>>> www.ossim.org
>>> Ossim-developer mailing list
>>> Oss...@li...
>>> https://lists.sourceforge.net/lists/listinfo/ossim-developer
>>
>
>
>
> -- 
>
>
>
>
> ---------------------------------------------------------------------- 
> ---
> SF.Net email is sponsored by: The Future of Linux Business White Paper
> from Novell.  From the desktop to the data center, Linux is going
> mainstream.  Let it simplify your IT future.
> http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4
> _______________________________________________
> www.ossim.org
> Ossim-developer mailing list
> Oss...@li...
> https://lists.sourceforge.net/lists/listinfo/ossim-developer