I have MODIS NDVI data of last year. I want to do crop type classification. Can anyone please suggest me how to do crop type classification by using these data???
Hi
The main aspect to point out is that you should use time series analysis since vegetations have different phenological behavior.
A further more subtle point is that you should try to select pure training pixels for each class. To this end, select big farmlands/orchards, preferably the middle of polygon to get ride of mixing problem and MODIS pixel size issue!!
There are different classifier and all them has their own advantages and disadvantages! In my view, SVM and RF are better than other. You can get use of feature selection in order to reduce data dimensionality. In your case, some data date may be redundant and consequently reduce classification accuracy.
In this way, you may draw acceptable results.
Good luck!
As mentioned either using the MODIS spectral reflectance of each crop or you can integrate the calendar dates of planting each crop with the NDVI values (where the highest NDVI values will be in the middle growth stage).
You can also check the PhD thesis.
https://www.itc.nl/library/papers_2011/phd/khan.pdf
It will better if you have GPS point of various cropland. Use these point to analyse the phenology of different crop. Make an NDVI threshold of each crop from series of data. Then use these threshold value in decision tree classification.
If study area has mono crop than MODIS NDVI will ok. But area, where heterogeneous crop practice is used then it, will not ideal.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs