[Qgis-user] Placing group of image at once

[Greg Troxel]

Léopold Stoessel via QGIS-User <qgis-user at lists.osgeo.org> writes:

> Our team is currently working on a project to reconstruct the railways
> of Canada using QGIS. To do this, we collect data directly on the
> rails, taking a photo every 2 meters. Each image (JPG) is accompanied
> by an XML file containing its GPS coordinates (EPSG:4326) and
> orientation.

Is the data really in EPSG:4326 (WGS84, the ensemble)?  Could it perhaps
really be in NAD83(CSRSv8) or some similar coordinates?

If it really is WGS84, and you collected it recently, you might label it
with the actual realization of WGS84 that was in use.  This will avoid a
definitional error of about 2m.

If you were using WAAS when collecting, then the data is in the WAAS
frame, probably ITRF2014, but I haven't been able to find where WAAS
documents that.

Unless you are using RTK, I am skeptical that your images are where you
think they are.  Cananda is big and much of it remote so unless you are
setting up local bases I don't see how that would work.

> Our goal is to display all the images on a QGIS map at their correct geographic location and with the proper orientation. (Example image: Try.jpg)

> As far as I understand, a JPG file needs either an associated PGW world file or must be converted into a georeferenced TIFF in order to be correctly placed on the map.
> Here are the challenges we're facing:
>
>   *   PGW method: I tried placing a PNG file with a PGW of the same name in the same directory, but QGIS seems to ignore the PGW file and creates HTML document with other coordinates instead.
>   *   GeoTIFF method: With over 500,000 images, manually georeferencing each one is not feasible.

There are georeferenceed image plugins.

For geotiff, that is for the situation where an image is oriented in the
plane of the surface, and one can relate pixel coordinates to ground
coordinates.  The standard example is aerial imagery.

For JPG, there are embedded metadata files "exif", and these include
location.  This is used for "I took this photograph at this location".
This is different from geotiff in that it encodes the camera location,
rather than giving formulas to compute the coordinates of each pixel.

> So I would like to ask for your advice:
>
>   *   Are we on the right track with our approach?
>   *   Which method would you recommend in our case?
>   *   Do you know of any way to automate the image placement and orientation based on our XML data?
>   *   Would it be possible to handle this through the Python console or using PyQGIS?
> Any guidance or suggestions would be greatly appreciated.

It is a small matter of programming to read an XML file with coordinates
and an image and produce a geotagged image.   You'll need to find or
hire someone who can do this.

If you're willing to publish your dataset under CC0, so that everybody
can use it, then probably someone will volunteer to write the code.