Research was carried out to explore possibility of using support vector machines (SVM) to estimate Cd concentration from hyperspectral reflectance. Canopy spectral measurements from rice plants were collected using an ASD field spectrometer in the experiment sites. Soil samples and rice samples were collected
for chemical analysis of Cd concentrations. A normalization spectral pre-processing method was employed to improve performance of the estimation model. Wavelet transforms were adopted to denoise the rice hyperspectral. Estimation of Cd concentration was achieved by an SVM approach. Compared to the original
(undenoised) hyperspectrl estimation model, the SVM model based on wavelet transforms yielded promising results with a coefficient of determination of 0.867 4 and a mean square error (MSE) of 0.001 2. The results indicate that it is possible to estimate Cd concentration in rice using wavelet transforms and SVM. This study can provide technical support for large area monitoring of heavy metals stressed crops using hyperspectral remote sensing.