1

I am completely brand new to eCognition Developer and want to test out Object Based Image Analysis (OBIA) on a Landsat scene of one portion of the Chesapeake Bay. I have been unable to find any online resources for beginners to learn the program or how to even begin doing this. Before going any further, is running OBIA even feasible on a medium spatial resolution image like Landsat? If not, I will continue to refine my pixel based classifications.

Note that I only have the trial version but I do have access to all available tools in the program.

1

I would highly suggest you joining the eCognition Community at:

(http://www.ecognition.com/community)

This is a great place to start with anyone that is new to the software package. On there people upload presentations, tutorials and rule sets to help with a wide array of problems.

With regards to OBIA for a Landsat scene, this is a great question and one that cannot be answered fully on here. Phiri and Morgenroth (2017) explain in great detail this very subject and find in their review that most researchers favour OBIA over pixel based-methods when using Landsat scenes.

However, you must bear in mind what you are trying to achieve with this. As long as your classes are broad enough then there is very little reason why a pixel based method would be superior. Image segmentation may struggle to clearly define boundaries, but this can vary greatly when it comes down to the segmentation process used within eCognition.

Link to paper: http://www.mdpi.com/2072-4292/9/9/967/pdf

  • thank you for the helpful comment! I actually read that paper previously but was hoping for maybe some more insight which is why I posted here. My land classes are beach, forest, weltand, urban, agriculture/open field, and water. Would these classes be considered broad enough for OBIA? – Paul M. Nov 1 '17 at 18:01
  • Those land cover classes/uses are fairly broad... perhaps even too broad. From my experience using OBIA to distinguish between wetland (which cover a wide breadth of land cover types) and forest land cover can be challenging, especially when only using optical/infrared imagery. It can also struggle to distinguish between highly photosynthetic wetland classes such as macrophytes and highly productive agricultural fields. – AWGIS Nov 3 '17 at 16:08

Your Answer

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

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