I'm working on a small project to extract water features from satellite images. In particular, I plan to use Landsat 7 ETM+ satellite images of a particular lake (it's on plain land). I have decided to use Decision Tree and SVM classifiers for classifying water and non-water pixels.
I am quite to new to remote sensing, and these might be trivial questions!
- Is there any standard way to get training datasets for the same? I do not plan to annotate the data myself. Do I train the data on some other (already labeled) images that has land/water regions?
- It'd be helpful (as a side question) if you could let me know how these training datasets are created?
- I also came across GSW dataset. Can it be helpful for this project in some way (for creating training datasets, or as a separate satellite image dataset)?
- Do people run a preprocessing step where they use known indices such as NDWI, MNDWI or AWEI, to capture the water pixels?
I also have access to ILWIS 3.3 software.