27

I have a non-georeferenced vector layer that I need to be georeferenced. With raster layers the task is easy and straightforward, but I have no idea what should I do with my vector layer. I have a few control points with known coordinates which should provide some basis to transformation. So, let's say I know points with id-s of 1, 2 and 3 should have the coordinates of x1,y1 ; x2,y2 ; x3,y3. There might be some rotation and scale transformation in addition to simple shifting.

Any ideas?

  • Have you tried the qgsAffine plugin? Similar question here: gis.stackexchange.com/questions/22691/how-to-georeference-a-dxf – Rayner Sep 11 '12 at 11:25
  • The Affine plug-in will probably do the transformation, but you will need to know the transformation parameters to use it. Can you post the vector coordinates and the corresponding real-world coordinates for your three points? N. – nhopton Sep 11 '12 at 11:38
  • Can you convert the shapefile to a tiff using gdal_rasterize, georeference the tiff, then extract the parameters from the world file? – klewis Sep 11 '12 at 20:55
10

To georeference a vector layer, try the qgsAffine plugin.

There is more info at Where to find qgsaffine in the menu?

  • 3
    Thank you for answers. I guess I have to use the qgsAffine then and figure out the transofrmation parameters manually. I was just hoping that there's some way to automate that process. Perhaps I'll try writing some Python code to do that – AHaav Sep 11 '12 at 11:46
  • 1
    Thinking about it, you might find that GRASS v.transform (it's in the Sextante toolkit) is friendlier to use than the qgsAffine plug-in. N. – nhopton Sep 11 '12 at 12:03
  • 1
    By the look of it v.transform will do the sums for you as well, see grass.osgeo.org/gdp/html_grass63/v.transform.html – nhopton Sep 11 '12 at 12:28
8

Given the fact that you have some points of control, you should be able to use an Affine transformation to shift your vector data. Have a look at this recipe. The process is a two part process:

  1. Use your control points to define the coefficients of your affine function required
  2. take the coefficients and apply them to the ST_Affine() in postgis.

If you put your control points into a CSV file (old_x,old_y,new_x,new_y), you can just about cut'n paste the R commands from the link to solve the coefficients part.

6

I recommend the Vector Bender plugin for QGIS. I tried it and it works fine and is user friendly. Depending on amount of pair of points that you define, you can either do:

  • translations: translation from one starting to ending point (1 pair)
  • uniform: translation, scaling and rotation (2 pairs)
  • bending: additional deformation (3 pairs or more)

You find a short video here and I advice you to read the Vector Bender help once you installed the plugin.

  • I agree, Vector bender plugin is intuitive and fast to work with. However, I could not manage to mirror/flip vectors. For this operation I use Affine transformation plugin (QGIS v2.18). – jurajb Jan 19 '18 at 22:19
5

With the release of GDAL 1.10 this is now possible.

It is done by performing a basic transformation in ogr2ogr and supplying the control points in the -gcp tag. http://www.gdal.org/ogr2ogr.html

For a guide please see: http://gisforthought.com/georeferencing-vector-data-using-qgis-and-ogr2ogr/

3

See my answer to Snapping two polygons together and give the free openjump a try.

I think that its affine transformation is much more user friendly than the qgis affine plugin.

2

I just had to do that, and ended up doing this :

  1. Rasterize the shapefile
  2. Georeference the raster using the Georeferencer plugin
  3. Save the GCP as a gcps.points file
  4. Compute the affine transformation using this file
  5. Apply the affine transformation to the shapefile using qgsAffine

The following script computes the affine transformation matrix using the saved GCPs :

# Computes an affine transform based on QGis GCPs
# Usage: gcp_affine.py gcps.points

import csv
import sys
import numpy as np
from skimage.transform import AffineTransform

u = list(csv.DictReader(open(sys.argv[1], "rb")))
source = [(d["pixelX"], d["pixelY"]) for d in u]
dest = [(d["mapX"], d["mapY"]) for d in u]
source = [map(float, s) for s in source]
dest = [map(float, s) for s in dest]
source = np.array(source)
dest = np.array(dest)
aft = AffineTransform()
aft.estimate(source, dest)
np.set_printoptions(suppress=True)
print aft._matrix
2

Following-on from my comments on Rayner's answer, GRASS v.transform, which can be run from the Sextante toolbox, can be used to calculate transformation parameters and apply them to a vector layer to perform an affine transformation. A text file containing control points is required, in the format shown here.

It's very easy to use and works well.

  • Maybe I am getting old, but just wondering whether folks remember the ShapeWarp user addon in Arcview 3.0. It was so easy to use. I have always found these other tools (v.transform/ affine) much more cumbersome. Hasn't someone tried to recreate the ease of ShapeWarp on QGIS? – Sharad Mar 19 at 6:25
-1

You can use the spatial adjustment method in ArcGIS Desktop. Once you add into arc Map, you need to make it editable than after spatial adjustment option is enable.And you also import your control point Spatial Adjustment->Link->Open Link.

For more details please refer Spatial Adjustment section of ESRI help

  • 3
    The asker had tagged the question with qgis. So I assume he does not have an Arcgis license. – AndreJ Sep 29 '14 at 11:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.