Declination and Ground Control

[mwillis]

I realize that Photoscan Pro does not presently use yaw, pitch, and roll angles when calculating photo positions from Ground Control.

Can anyone confirm that Photoscan Pro does use declination (orientation of the photograph relative to north)?

Thanks!

[bisenberger]

PhotoScan Po projects are orientated to the coordinate system your ground control points are in. It supports local coordinates, geographic coordinates, and projected coordinates (such as State Plane and UTM).

[mwillis]

bisenberger, thank you but that is not what I'm asking.  Let me try rephrashing the question:

Is the rotation of the photograph, in relationship to north, considered during alignment of photographs when the Ground Control option is used?

Thanks,

Mark

[bisenberger]

Hi Mark,
This image shows a project that uses control points for a local coordinate system. The axis gizmo in the lower right indicates the project orientation. The Y axis corresponds to grid north. The blue rectangles indicate the camera/photo locations and orientations.

[mwillis]

bisenberger, I think we're talking past each other.  I have flown many aerial photography missions and know the basics of Photoscan.

I'm curious if Photoscan considers the orientation of the photographs (i.e. looks at the EXIF header to see which way is up for each photograph) and then uses that data to help match it to an adjacent photograph that might be rotated 180 degrees.

I am testing a ground based mapping system that uses transects and a camera pointed down.  I will add orientation information to the EXIF header if it will help Photoscan during the alignment step but if it isn't used, I won't bother.

Thanks and I really appreciate your taking the time to respond.

[RalfH]

mwillis,

I can't say for sure whether Photoscan would use camera orientation data if they were present, but I doubt that it would make a difference. Alignment works almost exclusively through matching feature points (probably with some guidance from markers if they are present) and does not need camera orientation.

Your ground-based mapping system sounds very interesting though - are you talking about walking around in an area with the camera clicking along, taking measurements or samples along the way and later using Photoscan to reconstruct camera positions and sampling locations?

[mwillis]

Ralf, that is exactly what I'm talking about.  Below are a couple of test images:


I used a simple and cheap Garmin GPS, running a track log in my pocket while I took photos facing mostly down and a little forward as I walked across a large horizontal petroglyph site.  Later, I synced the photographs to the GPS track log and exported out new JPGs with latitude and longitude EXIF headers.



That allowed me to generate an oriented model with a darn good scale.



close-up


I need to map a giant area in this fashion.  The resolution of my UAVs cameras' isn't high enough to get the detail required.  The example above is only two transects and a test of the concept.  Promising...

-Mark

[RalfH]

Yes, this is pretty much the way I do it; I am either simply walking with the camera in my hands or use a pole (up to six metres) with the camera automatically shooting every few seconds (using CHDK).

I have done some tests using two GPS loggers simultaneously and comparing the results. Usually the models/orthophotos correspond reasonably well (especially when using many hundred individual positions), but sometimes standard GPS can be problematic (especially for elevation and especially for long, narrow projects like the one in your images).