> I am guessing this image is actually in Australia, in a rural area > "Alice Creek Nature Refuge", west of Brisbane?
Yes that's correct. > I would suggest talking to the people who published this and asking > them for the CRS. At least trying. I find it very puzzling that > it's close to 3857 but not really. I'd expect some UTM in some > version of GDA, or some conventional grid that's TM or Lambert > Conformal Conic or similar, analogous to the US "State Plane > Coordindate System. But maybe it's published for normie web mapping. The government spatial data is usually either GDA94 or GDA2020, but this is from QImagery, a historical archive of old aerial photography, so I can only guess that they were roughly assembled by people who are not GIS experts and likely just accepted the defaults. Many of the images are only approximately geotagged, and some are not even rotated to the correct orientation, so I don't think they are too interested in fixes as it looks like it's all best effort only. I have contacted them as I found a few images quite significantly off so we'll see what they say if they respond. > When you say "without the jgw", do you mean you put the JPG file in a > directory (without the jgw file in the directory!!) and then try to > add the file in the georeferencer? You can do this without having > the file as a layer - that's what I've always done. Yes, although I just moved the .jgw into a different folder. That allowed the transform to progress and I got a GeoTIFF out of it, but the coordinates were way off and didn't even remotely match the points I'd mapped. This only seems to happen with the colour imagery, the older aerial photography is black and white and those ones georeference ok (with the .jgw file present) , but again the points don't match the points I selected on the map very closely at all - a few hundred metres out. I ended up just giving up and moving my mapping points a few hundred metres in the opposite direction, re-referencing, tweaking the points, re-referencing, until the landmarks in the images appeared at the correct spot, with my georeferenced points bearing little resemblance to reality. So the "source CRS" does seem to influence the georeferencing process somehow, it's not as simple as mapping a bunch of pixels to points on the map, like I originally thought. I have geomapped raw images before and it worked fine, mapping specific pixels to points on the map (with a little wiggle room of course) but I've never before seen it so grossly ignore the points I mapped. > I also suggest trying the same thing with 3.44.x, on a different > computer if necessary to avoid messing up your installation. I switch > back and forth to test the in-progress qgis4 packaging, but I'm using > pkgsrc on NetBSD and I don't know what OS you are on, so that may be > harder. I might try it if I find more files - for now I've manually exported/reprojected/geomapped all the images I had, and it's a tedious enough process that took most of the day so I'm not too eager to repeat it :) > The jgw file is odd. [...] You could try with a jgw file with those > modified to 0.0000000000. It may well fix the export problem, but presumably the issue where the georeferenced points don't line up with the points selected on the map would remain - since that problem still exists even when the .jgw is removed. There must be some georeferencing information still present in the .jpg file? I have no idea, none of it makes a whole lot of sense to me! Cheers, Adam. _______________________________________________ QGIS-User mailing list [email protected] List info: https://lists.osgeo.org/mailman/listinfo/qgis-user Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user
