Quantification of Total Phenol in Grape by Near Infrared Spectroscopy and Artificial Neural Network

Abstract

Grape is one of the most important fruits in the world. Phenolic compounds are antioxidants are important compositions of grape. Phenolic compounds phrase includes all the aromatic molecules consisting amino acids to complex molecules like tannins and lignin’s. Near infrared spectroscopy is one of the most common nondestructive methods for fruits and vegetables qualification analysis. This research is conducted to evaluate the possibility of the quantification of total phenol in grape by near infrared spectroscopy and artificial neural network (perceptron). The number of 444 samples (107 Asgari, 106 Bidane, 111 shahroodi and 120 khoshnav varieties) were selected to model calibrating and test as well. Developed ANNs were compared on phenol prediction by residual prediction deviation (RPD) index in the test sample dataset (101 samples).The maximum RPD was 1.66 by 8-5-1 topology with correlation coefficient and root mean square (RMSE) equal to 0.79 and 48.66 respectively. It was concluded that NIR spectroscopy and back propagation perceptron ANN could be used to discriminate low and high amounts of grape total phenol as a nondestructive method.

Keywords

• grape
• neural networks
• nondestructive
• spectroscopy
• total phenol