I'm exploring several methods of classifying multi spectral data in eCognition.

The classification will be fairly high level; I'm not trying to find individual species but want to identify areas of forestry, urban areas and water bodies.

Most methods require training areas, but in this instance, I need an automated method. eCognition appears to be the best solution as you can create rules based on the spectral characteristics of the land cover types but this is an expensive solution!

Has anyone used any alternative methods?

  • I think that eCognition is the best one. However, necessarily, you will have to do visual inspection to correct some errors in classification result. Commented Sep 7, 2016 at 12:31
  • Thank you for the input Diogo, yes while developing the rulesets visual inspection will be needed, and having used eCognition in the past, this seems to be the most user friendly way of doing this. Once developed fully though the objective is that the result won't have to be checked so thoroughly. Commented Sep 7, 2016 at 16:01
  • 1
    I wrote a blog article on the subject a while back. In the example, I extracted water features, but you apply the same principles to extract other classes too. redfoxgis.com/single-post/2015/06/28/…
    – Aaron
    Commented Sep 8, 2016 at 5:54

1 Answer 1


eCognition is specialized for object-based analyses. For your classification problem, where your level of identification is only attributed to the land use / land cover, I recommend you to use the free software SNAP by the European Space Agency (http://step.esa.int/main/toolboxes/snap/). It provides some tools for unsupervised (no training data needed) classification algorithms like k-means, which is perfectly able to distinguish between land cover classes in multispectral data.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.