[4-1600-C] Postharvest/Food Technology and Process Engineering (4)

[4-1600-C-04] Quality and Shelf-life Prediction of Fresh-cut ‘Phulae’ Pineapple by Using Image Analysis and Artificial Neural Networks

*Rattapon Saengrayap¹, Mayura Dongsuea¹ (1. Postharvest Technology and Logistics Program, School of Agro-Industry, Mae Fah Luang University(Thailand))

Keywords:Browning index, Fractal dimension, Image features, Image processing, Storage

The image analysis technique had been applied for determining quality of fruit as a low cost, fast, and effective technique. In this study, the extracted image features, i.e., color, size, and texture, were then used as a criterion for predicting fresh-cut fruit quality and shelf-life. The aim of this study was to develop the suitable artificial neural network (ANN) model for predicting quality and shelf-life of fresh-cut ‘Phulae’ pineapple by using the information from image analysis. A green-yellow maturity stage of ‘Phulae’ pineapple [Ananas comosus (L.) Merr.] were cut into a cubical shape of 2×2×2 cm³ and stored at 5 and 10°C for 0, 2, 4, 6, and 8 days. The color (CIELAB values), shrinkage coefficient and firmness of fresh-cut ‘Phulae’ pineapple were determined every two days. The results showed that a higher storage temperature had a strong influence on the change of color. The L* values decreased since browning occurred resulted in the increased of browning index (BI). Moreover, the greater storage temperature also influenced the higher change of fruit shrinkage and firmness. The texture of the fresh-cut pineapple became softer and shrinkage was obviously observed as storage time increased. On the other hand, the image analysis technique was also used to assess the quality of fresh-cut ‘Phulae’ pineapple. The change of RGB color values, fruit dimension, and fractal dimension (FD) value were determined. The results showed that the R (red) value increased as the browning occurred. Moreover, a large variation of browning intensity and its area on the pineapple surface resulted in a larger FD value. The multi-layer feed-forward back propagation ANNs were developed to predict quality and shelf-life of fresh-cut ‘Phulae’ pineapple. The inputs of the model were storage temperature, fruit dimension, R and FD values, while the outputs were BI, shrinkage coefficient, and shelf-life. The numbers of the hidden node in a hidden layer were varied from two to forty with the increment of two. According to the selection of the best model for predicting the quality and shelf-life, the 18 hidden-node architecture was the most suitable model which provided R² of 0.98 and mean square error (MSE) of 0.01, 0.002 and 0.2 day for BI, shrinkage coefficient, and shelf-life, respectively.