The mean sugar content of a whole bunch of grapes and the mean sugar content distribution pattern by parts were
investigated in order to determine the NIR scanning area for evaluating grape sugar content. The coefficient of
determination between the mean sugar content of 20 berries (10 berries from the upper part and 10 berries from
the middle part) and the mean sugar content of the whole bunch of grapes was 0.9038 and, the maen sugar content
of 20 berries was close to that of the whole bunch of grapes. Thus, if prediction model is made for the non-destructive
grape sugar content by the NIR spectroscopy by acquiring each spectrum of the upper part and the middle part,
this sugar content representing the that of whole bunch of grapes. The sugar content evaluation model was made
by using the PLS regression. As a result of conducting the sugar content evaluation test on campbell early grapes
grown in a vinyl house, the standard error of calibration(SEC) was 0.40 brix, and the sugar content predictive accuracy
(SEP) was 0.48 brix. As a result of conducting the sugar content evaluation test on campbell early grapes
grown in outdoors, the SEC was 0.51 brix, and the sugar content predictive accuracy (SEP) was 0.62 brix, which
was a little higher than the grapes grown in the vinyl house. Meanwhile, in order to investigate the reproducibility
of the measured sugar content data, the sugar content of the same sample was repeatedly measured 3 times. As a
result, the coefficient of determination was 0.8685 and 0.8457, showing that the repetitive error was small. The
mean standard deviation was 0.21 brix.The mean sugar content of a whole bunch of grapes and the mean sugar content distribution pattern by parts were
mean standard deviation was 0.21 brix.