I have a drone image is which is georeferenced over a Google Sat. image. These align well, and both use WGS84-Pseudo Mercator (EPSG:3857).

I have a set of data points in a shapefile which are provided as WGS-84 (EPSG:4326); and QGIS is set for on-the-fly re-projection. Overlaying the data points over the images, I notice that all the data points are offset from the underlying physical objects seen on the images -- for the most part the shift is consistent across the entire data set of 20 or so points (allowing for GPS errors at the time of data collection, I suppose)

My question is this:

How does one physically shift the data set, that is by shifting the CRS to a custom CRS, by that small offset?

I would assume that I create a custom CRS, based on the existing WGS84, and then add some terms to the command line which accomplish the shift in the X and Y directions, by the 10-20 feet. However, I can't find an instruction tool/set online that explains this. Help...

Alternatively, I would have hoped for a way to merely enter a vector to define that shift (that is, to create a vector from a data point to the what should be the underlying physical object in the drone photo)... but I don't see that capacity in the software.

Yet another alternative is that there is some odd/weird CRS in the data collection which resolves the shift, and/or a weirdness in the WGS-84 pseudo-Mercator projection around where my data lives on the ground. So, I wonder if there is a way to discover which other formally named projections in QGIS live within 10 feet of my data points (and therefore by using that existing alternate, I get my data points shifted).

Of course, I can't disallow that the underlying Satellite image is, as a base map, shifted. In my use, in this situation, I could shift the GE Image, using a similar CRS adjustment. So... same questions.

I am using QGIS 2.6.1. I realize I could drag each individual data point, manually. Part learning curve, with this question, in advance of a task that will have far more points.

  • 1
    I wouldn't try to move the CRS - its conceptually wrong. Can you use the field calculator to define a "corrected point" for each of the observed points?
    – BradHards
    Commented Dec 27, 2014 at 22:04
  • An interesting assertion: why "conceptually" wrong. The change, using a field calculator shifts the space for the entire layer, just as the CRS change would; only, with a field calc the projection change is hidden within the scope of the properties dialog. As it happens, for all I know the data points are in some unknown projection, as collected originally. Still, the learning curve on the CRS changes contains documentation somewhere that I can read. But where...?
    – John
    Commented Dec 28, 2014 at 7:31
  • Consider: what happens when there are more points, this time without the same bias? If you move the points this time, you're correcting the data. If you change the CRS (to something that "looks right" but is really an offset from "mathematically correct") then you're biasing off future data. If the original data isn't in WGS-84 lon/lat, find the right (real) CRS and transform the data.
    – BradHards
    Commented Dec 28, 2014 at 8:03
  • I agree with you. This set is a fixed set and will always live on a layer. I would still like to understand the sytanx of the CRS adjustment. As an aside, I could never know if the Dec 2013 Google Earth is correct at that location, or shifted, and perhaps I should shift the CRS for GE Layer. Same issue. Same Learning curve.
    – John
    Commented Dec 28, 2014 at 13:49
  • 1
    I would first make certain that it's actually the vector data which is wrong and not the image and the Google sat image. It's not unheard of that Google sat images via OpenLayers plugin would be off by some distance.
    – underdark
    Commented Dec 28, 2014 at 20:38


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.