Abstract
Maize field yield and feed composition contents at the milk-ripe stage were estimated using multivariate analysis of hyperspectral (radiance spectral, first-derivative spectral) data at the yellow-ripe stage obtained through aerial remote sensing. Results show that when radiance spectral or first-derivative spectral data in the observed wavelength range (400-1,000nm) were used, R2cv between the measured value and the estimated value was a maximum of 0.18. The evaluation index (EI) rank was D. For EI of more than C, the method can be inferred as sufficiently accurate for practical use. With a selected waveband, however, R2cv was a maximum of 0.72 (P<0.05) and the EI rank was high at B. Furthermore, for a selected waveband, better improvement in estimation precision was achieved using first-derivative spectral data than when using radiance spectral data. Applying waveband selection and the first-derivative to spectral information observed from aerial remote sensing is an especially useful technique for aerial remote sensing where soil exposure and mixed pixel signals present difficulties. The yield and ether extracts (EE) increase at a fixed rate from the milk-ripe stage to harvest time, although crude protein (CP), organic cell wall (OCW), and crude ash (CA) decrease at a fixed rate. Results show that the spatial distribution of yield and feed composition content at harvest time can be estimated using these hyperspectral data at the milk-ripe stage.
