enter image description here

My goal is to find house sketches in archaeological data. I have a shapefile from an excavation where all the archaeological features, mainly pits are measured as polygons. The polygons have different sizes and shapes, but are mainly roundish.

The assumption is, that houses in prehistory consist of pits that stand in line and have roughly the same distance to another or make up a rectangle or are symmetrically arranged.

Is there a way to highlight pits/features in the data that match the above criteria so that e.g.

  • All the features that build up a rectangle or construct a right angle are marked or the rectangles/right angles are drawn to another layer?
  • All the features that are symmetrically arranged are selected or marked in the same colour?
  • All the features where the distances in between the features are similar (e.g. in between a range form 180 cm to 220 cm) are selected?

I cleaned the Data and selected some features that are likely to be Postholes. After that I started to change the features with the tool centroid to points. The idea is now to connect all the dots with lines and select the Houses via the length of the lines. I tried to connect the dots buy using the Delaunay triangulation and convert the polygons into lines afterwards. Sadly I don't get only single lines but also i have lines that have still a triangular form.

Is it possible to convert the polygons in a way that I have only lines without the vertices?

  • 1
    This is a very interesting project. Before proceeding, however, I think that you need to define what "symetrically arranged" means to you, and establish a priori what geometric arrangements you will look for.
    – Stu Smith
    Feb 19, 2021 at 16:15
  • 2
    Can you send the shapefile? If you can't, please add a picture to the question.
    – Xi Jin
    Feb 19, 2021 at 18:57
  • Thanks for your comments so far. @StuSmith I think that will be one of my problems because archaeological data shows, that even symetric houses have varying distances between posts. E.g. if I have a rectangle house out of 6 posts that has something around 6 m by 4 m the distrances between posts vary around 3 m/4 m depending on the axis. I think i have to use ranges of distances e.g. like from 2,75 m to 3,25 m. The other thing is, i have to take the diameter form the feature into account. Maybe i can use the geometrical center of the postholes/polygons and generate a pointlayer with that.
    – lqdxtc
    Feb 21, 2021 at 21:14
  • I think that setting up anything based on an imperfect dataset will lead to false positives and negatives. If we are talking about millions of feature you could just filter by size and use a buffer to see groupements. But I don't think there is a simple answer that will be faster to implement than doing the work 'manually'. Other data sources may be helpful.
    – Al rl
    Feb 21, 2021 at 21:50
  • This article may help you avoid bad decisions: badscience.net/2010/01/voices-of-the-ancients
    – wingnut
    Apr 23, 2021 at 11:33


Browse other questions tagged or ask your own question.