Hyperspectral remote sensing estimation models for nitrogen contents of maize

The hyperspectral reflectances of different organs(including leaves, stem) and corresponding nitrogen contents of different species of maize were measured in the experiment. The correlation among raw hyperspectral reflectance, hyperspectral characteristic variables, RVI, NDVI and chlorophyll contents were analyzed. The results show that the raw spectral reflectance has the maximum negative correlation coefficient at 716 nm(r=-0.847) with nitrogen contents and the logarithm model constructed with reflectance at this point is the better one as compared to linear model; the first derivative spectral reflectance has the maximum positive correlation coefficient at 759 nm(r=0.944) and the linear and non-linear models have the similar capacity for the nitrogen estimation; as to hyperspectral characteristic variables, they all show significant correlation with nitrogen contents except λy, SDr/SDy and (SDr+SDy)/(SDr-SDy). Those results indicate that the variables adopt in this paper have great potential for nitrogen content estimation. By the precision evaluation of estimation models, the logarithm models constructed by the first derivative reflectance at 759 nm was proved to be the best for the estimation of maize nitrogen contents.