Hi, 

I think the main difference and problem, comparing a features space of spectral 
bands and one of NDVI values in the time domain is that the spectral bands do 
not shift, but a NDVI pattern does.

I.e.
Wheat in different climatological areas (mountain/valley, north/south, 
continental/...) will have (practically) same NDVI patterns, only shifted in 
the time domain because of climatological and other reasons.

I don't know any standard functions in R, but try with Correlation, RMSE and 
Spectral Angle. To reduce the shift in the time domain, it's possible to move 
the two curves (reference data vs. pixel to classify) by shifting them stepwise 
in the time domain, and then use the best fit for the classification.

I hope this helps a little, let me know if you find something better! I'm 
interested too.

cheers, Matteo


>>> Alexandre Villers  11.06.11 10.45 Uhr >>>
Hey,

 From what I understood, most landscape classification methods available 
from GRASS (through R) do not account for the temporaI aspect of 
spectrum. So for example, if you have a raster where each band 
represents NDVI values at a given date, from March to September for 
example, the classification will not consider that values are 
time-dependant. I would like to use time series of NDVI values to 
classify pixels in different type of crop, accordingly to the shape of 
each time series (rapeseed will not present the same NDVI time serie as 
would alfalfa or wheat).
I was wondering if someone could point me to a method that would allow 
such classification using R ? I started looking at Functional Data 
Analysis but I must admit that I'm a bit lost at the moment.

Thanks for any help


Alexandre

_______________________________________________
R-sig-Geo mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo


        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo

Reply via email to