Is it possible to georeference this type of angled cross section, preferably in QGIS (Section-II in the image) ?

I am very much new to this type of georeferencing.

Googled & couldn't find a proper lead

enter image description here

  • Great topic! My favorite since university times. There seems to be no significant progress in supporting geologist in this kind of basic geometric transformations. What do you want to do with georeferenced images? Digitize in XYZ-space? Reproject to 3d-space? Interpolate or triangulate geologic surfaces? Visualize a network of cross section? Thus the answers can reach from YES - we can, a set of trigonometric formulas for the transformation, to NO - postprocessing in QGIS & co is to tricky.
    – Detlev
    Dec 23, 2015 at 8:13
  • @Detlev Yes. We would like to digitize & export it as a 3d shapefile. Then visualize those 3d shapefiles in a 3d space. Could you provide me any leads regarding this ?
    – joseph_k
    Dec 29, 2015 at 8:34

1 Answer 1


I suggest to decompose the problem into several aspects.

First: Georeferencing images of cross sections

The image of a cross section is only an image, like others too. You can georeference it against coordinate pairs with depth and length components. Eg 0,0 in the upper left corner, or whatever the depth and position along the section line of this point is. This leads to an unfolded, exaggerated plane with Y corresponding to true z (depth), and X corresponding to some position in true x, y plane (ground).

This leads to the question: how to convert the X,Y image coordinates to x, y z. As said above, Y corresponds to z devided by the exaggeration of the sketch in the image.

Second: Transform digitized polylines to 3d space with x,y,z coordinates

To convert X to x, y you need x, y coordinates from startpoint and endpoint of the section line, and every vertex, where direction of section line changes. You already have this information, if you have digitized the section line. To bring all together you need something to link the section line to the digitized cross section. This could be modeled by simply adding an attribute to the feature class, holding the cross section, which contains the id of the section line.

What remains to do is very similar to linear referencing: pick a position along a polyline (position corresponds to X of the cross section), and return its x,y coordinate. To do this:

  1. get the section line

  2. for each vertex determine its distance from the preceeding vertex

    sqrt((x2-x1)**2 + (y2-y1)**2)

  3. save this distance along with the coordinates of this and the preceeding vertex in a dictionary with a tuple of the distances as key and list of 2 coordinate tuples as value

Now, do the transformation:

  1. get the appropriate dictionary (that corresponding with the attribute of the cross section features)

  2. for each polyline of the digitized cross section, for each vertex of the polyline

  3. get that dictionary entry having a key whose first component is equal or greater than X, and whose second component is equal or smaller than X as pos

  4. apply rule of proportion: get relation of X-pos[0] to pos[1]-pos[0]

  5. get the dictionary value (remember: coordinates of the endpoints of a straight section line segment), multiply the differences in x and y of both tuples with the relation from 4., and add the result to the first tuple (coordinate of segment start point)

Third: Generate 3d feature classes

This algorithm can be implemented as Python script which creates a 3d feature class, transforms every polyline, and appends it with some useful attributes.

These 3 steps are easy to work out with QGIS and some scripting.

Fourth: Visualization

To just visualize and explore 3d shapes (vector and images only), I prefer a CAD environment (SketchUp or Rhinoceros3d) which allows for much more interactivity and rendering capabilities then GIS. There are some threads discussing visualization techniques in QGIS/GRASS (here or here).


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.