M. Codrea, E. Tyystjärvi (Finland), M. vandeVen, R. Valeke (Belgium), and O. Nevalainen (Finland)
Fluorescence image processing, neural network classifier, geometrical features.
Classification of harvested apples when predicting their
storage quality is economically important. We describe
a new automatic apple classification system based on
chlorophyll a fluorescence images taken in blue light
through a red filter. Fluorescence appears as a
homogenous area broken by of small non-fluorescing
spots, corresponding to normal corky tissue patches,
lenticells, and to damaged areas that lower the quality
of the apple. The classification relies on geometrical
features of the spots and employs two neural networks.
Experimental results show high potential of the
technique. The system reached 96% accuracy in