|Lee, Kang-Jin - NATL AG MECH. RES UNIT|
|Choi, Chang-Hyun - SUNGKYUNKWAN UNIVERSITY|
|Choi, Kyu-Hong - NATI AG MECH RES UNIT|
Submitted to: ASAE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: April 2, 2002
Publication Date: July 28, 2002
Citation: Park, B., Abbott, J.A., Lee, K., Choi, C., Choi, K. 2002. Near infrared spectroscopy to predict soluble solids and firmness in apples. American Society of Agricultral Engineers Annual International Meeting. Paper No.02-3066. Interpretive Summary: Nondestructive measurements for predicting apple quality attributes are very important to apple producers and processors. Firmness and soluble solids are most important internal quality attributes to determine the consumer acceptance and shelf life. Visible/Near-infrared (Vis/NIR) spectroscopic technique could be used for apple quality measurement. The soluble solids and Magness-Taylor firmness of apples were predicted by Vis/NIR reflectance measurement. The soluble solids of both Washington Delicious and Pennsylvania Gala apples showed good relation with NIR diffuse reflectance data. Even though the accuracies of prediction models for firmness were much lower than soluble solids, the spectroscopic reflectance measurement of firmness was feasible for Delicious apples. Classification models performed well to classify apples qualitatively based on soluble solids for both Delicious and Gala apples with high classification accuracy. For the firmness classification, however, the overall accuracies were much lower than soluble solids classification for both Gala and Delicious apples.
Technical Abstract: Soluble solids could be predicted by Near-infrared (NIR) spectroscopic technique with principal component regression (PCR) and Mahalanobis distance (MD) analysis. The coefficients of determination for predicting soluble solids were 0.93 for Gala apples, and 0.96 for Delicious apples with only NIR reflectance measurement. For classifying apples into three classes based on the soluble solids, MD classifiers had accuracies of 93.5 percent for Gala and 92.1 percent for Delicious apples when Vis/NIR reflectance was measured. While, when only NIR was used, the classification accuracies were 95.5 percent for Gala and 93.6 percent for Delicious apples, respectively. However, prediction accuracy for firmness was low. Using PCR models, the coefficients of determination for predicting firmness (Magness-Taylor maximum force) were only 0.22 for Gala apples and 0.79 for Delicious apples with Vis/NIR; while, the coefficients of determination were only 0.29 for Gala and 0.65 for Delicious apple with NIR only. Based on three classes using MD analyses, the classification accuracies were 82.5 percent for Gala and 83.8 percent for Delicious apples when Vis/NIR reflectance data were used. However, when only NIR was used, the classification accuracies were 80 percent for Gala and 75.3 percent for Delicious apples.