Decomposition of mixed pixel for cotton identification using remote sensing data

In order to improve cotton identifying accuracy, taking Manas county in Xinjiang province as study area, the linear spectral mixture model (LSMM) was applied to the study of pixel unmixing technique based on TM remote sensing data. Four typical endmember spectrum values were put into the linear model, including spectrums of cotton, corn, tomato and soil. Under unconstrained condition, the mixed coefficient was derived by the least square method, together with the abundance of each surface feature and RMS error chart. The results of pixel unmixing were tested with ground measurement of the cotton field in the study area, which showed that the LSMM modeling was simple with less calculation, and the precision of the decomposition of mixed pixels exceeded 90% that's enough for cotton identification with remote sensing data in Xinjiang province.