Elektrine lite

← Feed

@mjambon@mastodon.social

2024-11-24 17:45 UTC

@tim_fan cool. I don't have any need for it but I'm curious about dealing with bad data in practice and in theory. I.e. if one point is inconsistent with the majority, there should be a way to detect it and flag it. In practice, I suppose it depends on the application (e.g. are coordinates entered by a human who can make typos?).

Replies (2)

  • @tim_fan@mastodon.social 2024-11-24 18:02

    @mjambon yeah I'm sure for my maps I make mistakes. All I do for error checking is to perform the forwards warping onto satellite imagery in QGIS, then visually check for discrepancies between the two. With this I can catch gross errors. E.g. overlay: https://www.facebook.com/photo/?fbid=315155968310211 Beyond that, perhaps you could, on a per point basis, leave the point out of the referencing process, then see where it gets mapped to. If it is mapped far away from its reference location, perhaps the reference is wrong.

    Open ##1974642

  • @mjambon@mastodon.social 2024-11-24 18:05

    @tim_fan I suppose we could compare the Delaunay triangulations in both coordinate spaces and flag neighborhoods that are not preserved by the mapping.

    Open ##1974643