Abstract
A 4.1MHz$1^H$ Nuclear Magnetic Resonance(NMR) sensor was designed and manufactured to evaluate the internal quality of fruits. The magnet console having 963gauss magnetic field induction was used for the NMR sensor. To optimize and evaluate the NMR sensor, glycerol and sugar-water solutions were used. $^1$H(proton) resonance signals were used to estimate the sugar contents in fruits. Artificial neural network models were developed to predict sugar contents in fruits from the proton resonance signals. The standard errors of prediction(SEP) were 0.565(apple), 0.394(pear) and 0.415(kiwi), respectively. The result implied that it was possible to evaluate apple, pear and kiwi into 3 grades using the NMR sensor.