Study of Applying Near Infrared Spectroscopy for Determinating the Sugar Content of Wax Apple Fruits

• Main Author: 鍾昭台
• Other Authors: Ching-Wei Cheng
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/81572073967109831660

碩士 === 國立中興大學 === 生物產業機電工作學系 === 91 === In this study, the near infrared spectroscopy (NIRS) was used to determinate the sugar content of wax apple fruits in the wavelength range from 500nm to 1000nm. In the experiment A, the wax apples in Pingtung Fungliao area were chosen as the fruits to be detected. By using non-destructive measurement technique, the total calibration curves were established. And in the experiment B, it was to find one method of amending calibration curves, which was used for different time, region or variety, and then to modify the determinative mode quickly and conveniently. By applying multiple linear regression (MLR) analysis, the correlation of sugar content and second derivative spectra could be obtained. From the results of experiment A, it showed that the total calibration equation which consisted of 5 wavelengths (i.e. 952nm, 642nm, 884nm, 906nm and 858nm) had the calibration group with =0.931, SEC=0.388, and the first prediction set with =0.937, SEP=0.262, Bias=0.309 and RPD=4.454. For the second prediction set, the statistic data were =0.964, SEP =0.207, Bias =0.251, RPD=5.551. Finally, the total prediction set had the results as =0.918, SEP=0.322, Bias =0.287, RPD=3.649. The results of experiment B showed that the calibration equation of the first group could have good ability to predict the sugar content of samples in the second group. The statistic data were =0.974, SEC=0.229 and =0.96, SEP=1.099, Bias=1.06, RPD=1.119. However, the values of measurement had 1 average deviation different with values of content from the prediction set. After amending the deviation, the results were SEP=0.267, Bias=0.267 and RPD=4.607. Obviously, the internal qualities of fruits would be different with different cultivating environment and sections. In order to reduce predicting error, it should be collect more samples to modify the calibration curve. To expand the measuring range, in the future, it has to establish a number of databases from different sections.