Vegetable growth stage estimation from an airborne hyperspectral image

We propose a new method to estimate growth stages of vegetables from an airborne hyperspectral image. The new method is based on spectral features generated from the spectral data of vegetables measured by field spectrometer, namely "field spectra". Field spectra show different curves depending on sorts and growth stages of vegetables. We selected spectral features based on relative and derivative values to identify the sorts and growth stages. We also introduced a hierarchical masking method to employ appropriate spectral features for each layer. To examine the fundamental performance of the method, we collected field spectra and took an airborne hyperspectral image of farms, where lettuce in different growth stages and other sorts of vegetables such as cabbage and Chinese cabbage were growing. It is shown that the proposed method can estimate the growth stage of lettuce, number of elapsed days after planted, accurately with small error of few days.