Is there a way to colorimetrically and automatically vectorize georeferenced rasters that came from manually colored, scanned maps? Currently we Print the topography maps 20k scale --> hand-coloring work (hazard mapping) --> georeferencing (perfectly, if I may say so) ---(current stage)--> digitization of manually colored map.

If a solution exists i know some eye-to-screen adjustment would still be in order though I believe this path would be easier in the long run.

  • 2
    this is almost like land cover classification - at least some raster re-classification can be done once you've scanned the map - how many colours are you dealing with? Once reclassified to 'just' those colours, converting to vector based should be easy... Jun 18, 2015 at 16:42
  • Yes, I have seen the r.reclass command. I'm just worried about satellite rasters vs crayon marks..I mean, how would the computer analysis fare.
    – wearashirt
    Jun 18, 2015 at 16:54
  • 1
    it's all pixels, right? Do you have a little snippet you can post of what the scanned image looks like? Jun 18, 2015 at 17:00
  • I have seen this done in ERDAS many (many, many) years ago with supervised classification on 50k and 100k printed topographic maps. Select your 'training areas' wisely! Other software would be able to do it (with varying degrees of success) but in the end the biggest factor is selecting your 'training areas' wisely - be prepared to do several iterations to get it right. Depending on the level of detail it should be quicker than heads-up digitizing. Jun 18, 2015 at 22:30
  • Sorry, I'm having trouble uploading a 1.4 MB jpg. Right now, I've merged the scanned, georeferenced maps in QGIS using Raster > Miscellaneous > merge. It's now in GRASS, where the RGB bands were apparently loaded separately, and all is B&W...
    – wearashirt
    Jun 19, 2015 at 1:13


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.