|
|
| |
| Application of Artificial Neural Network Modelling for ‘Phulae’ Pineapple Maturity Classification |
Habib Ullah, Johnson Ogunsua, Lutfullah Riaz Ahmed and Ameer Uddin
|
Abstract:
Classification method is one of the most important step for food quality evaluation using machine vision system. To replace human decision-making system, the accurate, concise and precise classification algorithm is concerned. This study aims to classify maturity of ‘Phulae’ pineapple using different artificial neural network modelling architectures and to compare the accuracy of maturity classification in the using of ANN model with the Partial Least Square (PLS) analysis. Pineapple fruit were harvested in three different season, i.e., summer, rainy and winter in the year 2017-2018. In each season, twenty pineapple fruits of each three different maturity stages (green, green-yellow and yellow stages) were selected for testing. Pineapple qualities which are RGB values, percentage of yellow peel area (%Y), TSS were evaluated. The results showed that percentage of yellow area increased as fruit aged. TSS also provided the same trend; TSS increased as maturity stage increased. In terms of RGB values, R value was the best color parameter to distinguish pineapple maturity by providing higher correlation coefficients (r=0.788, 0.925 and 0.847 for summer, rainy and winter season, respectively). All the experimental data were used for developing ANN models. Inputs of the model were RGB values, and %Y, while outputs were maturity stages. One-hidden-layer ANN architecture with the different hidden nodes (2-20 nodes) in a hidden-layer was tested. The results showed that the ANN with 20 nodes in a hidden-layer was the most suitable model for classifying ‘Phulae’ pineapple maturity with the highest R2 of 0.918 and lowest RMSE of 0.17 and 1.33 ºBrix for maturity stage classification. The PLS was also applied for classification with overall performance of 87.54%. In addition, R value and %Y were also found to be the main parameters to distinguish pineapple maturity. According to the results, ANN shows better capability to use as a classification algorithm for maturity classification in ‘Phulae’ pineapple.
|
|
|
| | |