抄録
Overripe berries cannot be classified from ripe berries of Japanese blue honeysuckle using RGB color imaging analysis since they have the same color appearances. Hyperspectal imaging analysis was thus employed. First, a pixel discrimination function was generated based upon forward stepwise selection to select significant wavebands (751 nm and 420 nm) and linear discriminant analysis, and it was applied to all pixels of segmented berries. Then a majority rule was applied to each berry independently for fruit object classification, and the classification was improved by eroding the fruit edge area. The classification success rates of pixel discrimination and fruit object discrimination tests were 69.7% and 73.9%, respectively. This study showed the potential of identifying overripe berries using hyperspectral imaging analysis.
