Hello everyone,
I am doing LULC of the arid region, I have acquired landsate 8 image data from USGS website and I have done preprocess in qgis using semi-automatic classification plugin using the standard tutorial and I have converted by DN into reflections value for LULC. however, I facing difficulties in assigning classes for built-up area and bare soil as they have high overlap spectral values.
besides this, I have also used SAVI as well as Modified bare soil index though it's not helping in my problem.
Anyone can tell me what to do in that case.
Thank you.
Usually, built areas are limited, and therefore, you may extract them manually by photointerpretation.
This is very common, as the spectral signatures b/n the built-up area and bare land (sandy area) are nearly similar. I recommend you extract built-up area indices. This link can help you Article Separating Built-Up Areas from Bare Land in Mediterranean Ci...
We have used Cubist and Deep Learning via Keras to do this same thing. You would need training data for these methods though. We also view the scenes using 6 band data with RGB being 643 or 642. For training the models we use all bands and some indices like NDWI, NDVI, wetness index, dryness index, brightness, greenness, etc.
I should add that at USGS EROS we also have NLCD Impervious USA data available and on the RCMAP side we have bare ground percentage for the USA. I work on both of these projects.
Addisu Damtew THANK YOU VERY MUCH FOR SHARING LINK, I will go through it.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps