In OBIA to extract landslide features, hillshade was used to removed false positives of shadow area but some true positives were also removed. There are some shadow areas in image containing landslides as verified from field. How can those features retrieved?
I dont know if there is a specfic feature extraction algorithme, but if there are not so many (Landslides), you can draw them manualy (digitalization), the convert the file to a binary-raster file then add them in eCognition as a "Thematic Layer" , and create this classe in the ruleset with : Thematic Layer = 1.
One of my former PhD students, Tapas Martha, for some years now a senior researcher at NRSA in Hyderabad did his PhD on landslide detection, and 3 of his thesis chapters (all published and highly cited ISI papers) deal with using eCognition. While the focus was not on detecting landslides in shadow, it was one problem he had to deal with. I recommend that you start by looking through his thesis/papers ( https://webapps.itc.utwente.nl/librarywww/papers_2011/phd/martha.pdf )
Good luck, Norman
- Badges
- Science topic
- Similar topics
- Computer Science
- Images