Normally ISODATA clustering is performed on the stacked time series of NDVI data. As a result we get land cover for different crops/vegetation in the study area labeled with class numbers. Now, how can we map these classes to the crops on the ground?
based on DN values we have to fix the classes (based on reflectance of NIR/IR Ratio)
With ISODATA you generate spatio-temporal NDVI profile, crop/vegetation phenology or cover phenology classes (all three the same, but labelled in a different terminology). You may vary the number of classes, you have no doubt experienced.
I suggest to relate a range of NDVI profile classes to the excellent agricultural census data available in Pakistan, although not in GIS format to the best of my knowledge.
@ Hein Van Gils yes one can relate classes with the crop areas but GIS data for such areas are not available in Pakistan. Alternatively, we can use crop colander data, but the problem here is to separate classes of the competing crops on same time line. Is there any statistical algorithm available to label the competing classes at this stage.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps