Orthorectification problems

1289
7
06-03-2013 02:51 AM
JohnMobley
New Contributor
Hello,

I'm setting up an image services that serves orthorectified aerial images.  A lot of the output looks good, but there are quite a few mis-alignments.  I don't see any consistency between the problem areas.  See attached screen shots for typical examples.

The image service uses and rpdef file to define camera interior orientation, and the image exterior orientation parameters are in a dbf that reference the rpdef file.  The service uses a DEM that I believe is accurate.  The inaccuracies are not necessarily close to changes in elevation, so I don't think that the DEM is likely to be the problem anyway.

I would appreciate any suggestions on how to debug the problems.  

I'm new to this, so I might be overlooking something obvious.

Thanks.

David

[ATTACH=CONFIG]24920[/ATTACH]

[ATTACH=CONFIG]24921[/ATTACH]
0 Kudos
7 Replies
larryzhang1
New Contributor III
Hello,

... quite a few mis-alignments.  ...

The image service uses and rpdef file to define camera interior orientation, and the image exterior orientation parameters are in a dbf that reference the rpdef file.  ....

[ATTACH=CONFIG]24920[/ATTACH]

[ATTACH=CONFIG]24921[/ATTACH]


David,

It looks that the model �??RAW to RPDef�?? should be re-done in PCI GI Tools firstly. You should combine GCP & TP with RPC to refine this model, at least, a few of TP (Tie Points) for your case.

After rewriting �??RPDefs�?? from GI, orthorectification can be done again with DEM within MD.

++++++++++
If GI tool not accessible (or not licensed), it is advisable to remove those mis-alignment images from the model 'RPDef'; and then use the tool �??Georeferencing�?? of ArcGIS to refine with TPs.
0 Kudos
by Anonymous User
Not applicable
Original User: johnmobley

David,

You should combine GCP & TP with RPC to refine this model, at least, a few of TP (Tie Points) for your case.



hlzang, thank you for the response.

I should have added that we're not using RPC's.  Our general process is to run the IMU data through Applanix POSPac MMS after the flight, generate an external orientation file, and use that as input to the orthorectification process.  We have not been using tie points or gcp. 

Is there a way to specify tie points in ArcMap?

Thanks.

David
0 Kudos
larryzhang1
New Contributor III
... to run the IMU data through Applanix POSPac MMS after the flight, generate an external orientation file, and use that as input to the orthorectification process.  We have not been using tie points or gcp. 

Is there a way to specify tie points in ArcMap?

...


David,

On your description, most likely, something more should be done, to refine this model during running the package of Trimble �??APPLANIX POSPac MMS�??, which may include the QA/QC processes such as with �??Applanix SmartBase module�?? to check �??Bad reference station data? Bad antenna heights? Bad reference station coordinates?...�??

Generally, this can be done quickly and easily by using a few overlapping images from each mission and checking the system calibration parameters using the automatic point matching and bundle adjustment. This ensures you achieve the highest quality results each and every mission.

Please check with the expert on this Trimble system and its package to help solve the model, as the best choice.

Not sure if PCI GI supports the camera type that you are using in IMU. Better to check with PCI. (If Trimple APPLANIX DSS used, perfect). If yes, the GI tool will most easily fix the model 'RPDef' directly with TP for your case, if licensed.

However, as a backup, as said earlier, you can remove those mis-alignment images from the model �??RPDef�??, and then perform the refinement with TP in ArcMap (�??GeoReferencing�??). After satisfied with the quality,  add those �??geo-corrected�?? images back to MD via Raster Type: �??Raster Dataset�??, which is combined with previously Raster Type: �??RPDef�??, for orthorectification in MD together �?�


Good luck,

larry @ ARAMCO


+++++++++++++

Personally, the system calibration was done not perfectly.

In fact, unlike Aerial Triangulation, Direct Georeferencing relies on the system calibration and direct EO to be accurate in order to achieve the final ground accuracy.

While the calibration of mapping cameras is stable, changes can occur over time due to a number of factors such as handling and age. Ongoing quality assurance and control of the calibrations parameters is hence crucial for success.
0 Kudos
by Anonymous User
Not applicable
Original User: johnmobley

Larry,

Thanks for the response.  That prompted me to call Applanix, and they pointed out something I was doing wrong.   It didn't solve the problems, but it did help in a few cases.

I'm seeing a couple of things that are confusing me.  One thing is that when I make changes to a small subset of the data, the subset looks good, but when I apply them to the entire data set, the changes don't do much.  I've attached a couple more jpegs that demonstrate what I'm talking about.  The  "nad83_with_subset_of_images" image shows how the change (switching to a nad83 datum in the EO) affected the subset (about 10) images.  The "nad83_with_all_images" image shows how the change did not have the same effect on the complete dataset.  It is slightly different than before I switched to NAD83, but it isn't helping as much as it did with the smaller subset of images.  Does that mean anything?

I'm also seeing that in ArcMap, when I zoom in more, the alignment gets better.  It seems to change the images it is selecting.  I don't see similar behavior in the deployed service (as a WMS), so I'm not sure what that means.

Any help you can provide is appreciated.

Thanks.

David



[ATTACH=CONFIG]24990[/ATTACH]   [ATTACH=CONFIG]24991[/ATTACH]
0 Kudos
larryzhang1
New Contributor III
...
The  "nad83_with_subset_of_images" image shows how the change (switching to a nad83 datum in the EO) affected the subset (about 10) images.  The "nad83_with_all_images" image shows how the change did not have the same effect on the complete dataset.  It is slightly different than before I switched to NAD83, but it isn't helping as much as it did with the smaller subset of images.  Does that mean anything?

I'm also seeing that in ArcMap, when I zoom in more, the alignment gets better.  It seems to change the images it is selecting.  I don't see similar behavior in the deployed service (as a WMS), so I'm not sure what that means.


[ATTACH=CONFIG]24990[/ATTACH]   [ATTACH=CONFIG]24991[/ATTACH]


david,

same as you and not sure why�?�

1. Two scenes have different datum from others (NAD83)? (It shouldn�??t be, because all are done in one project�?�);
2. Two scenes have real displacement against adjacent images. (Most likely �?�)
3. �?�

Honestly, your issue is really attractive to me.  Indeed, �??Direct Georeferencing�?? products look easy and simple to many end users; but, many practices show that it is intensive cooperation and detail-oriented mission�?�

Pls share, when you get it solved.
0 Kudos
by Anonymous User
Not applicable
Original User: johnmobley

david,

Indeed, �??Direct Georeferencing�?? products look easy and simple to many end users; but, many practices show that it is intensive cooperation and detail-oriented mission�?�



Larry

That is what I'm hearing from several sources.  It looks like it may require more manual processing than we expected.

Thanks for the suggestions.  I will let post when I have a solution.
0 Kudos
larryzhang1
New Contributor III
Just for information sharing,

10.2 will provide an effective mechanism to incorporate TP & GCP during creation of Mosaic Datasets... It is also a solution to your issue.

See more at: http://ideas.arcgis.com/ideaView?id=08730000000boIp#sthash.k4GndsqH.dpuf

Larry @ Aramco
0 Kudos