I have a panchromatic satellite image with a large coastline area and my goal is to delineate this coastline with a maximum accuracy and minimum amount of manual work. What is the best option to accomplish this?

I do not have either DEM nor other spectral bands, so these solutions do not work:

Here is a small area of the original 2-meter satellite image I have:

enter image description here

  • 2
    If that's all you have then you really have a serious problem. You could give supervised classification a go but I have my doubts that it will return anything like what you need. If you can source some infra red bands your chances of automatically classifying sea/shore edge get better. Perhaps you could start with SRTM to approximate your coastline and use that to limit your processing area to save some cleaning up later. – Michael Stimson Sep 3 '17 at 23:58
  • 1
    I would explore image segmentation with Python's opencv package. Here is a blog post that will hopefully get you started: learndeltax.blogspot.com/2016/02/…. Python's scikit-image package is also excellent for digital image processing: scikit-image.org/docs/dev/auto_examples – Aaron Sep 4 '17 at 4:55
  • 1
    I got a decent result thresholding then grouping low values . What software do you have available? – user2856 Sep 4 '17 at 5:16
  • @Luke you name it. What would you recommend? Your thresholding seems pretty good, although there are some brighter water areas present on this image, too – Basile Sep 4 '17 at 9:32
  • @Aaron thanks for the information about those packages, especially scikit-image! Do you have any experience of converting non-georeferenced outputs of those algorithms into georeferenced ones (i.e. detected contours to shapefiles)? – Basile Sep 4 '17 at 10:42

Your Answer

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

Browse other questions tagged or ask your own question.