Articles
Non-destructive methods for the prediction of the number and weight of pomegranate arils
Article number
1464_17
Pages
127 – 136
Language
English
Abstract
This study was aimed to non-destructively model the number and weight of the pomegranate arils.
In this study, ‘Wonderful’ pomegranate fruits were used as materials and the prediction of the outputs (number of arils and arils weight) had been done with five inputs namely fruit weight (g), fruit width (mm), fruit height (mm), fruit crown diameter (mm) and stem end diameter (mm). Predictions of the outputs were done with four different methods including multiple linear regression (MLR), principal component analyses (PCA), artificial neural network (ANN) and support vector machines (SVM). The findings of this study provide evidence for the ability of non-destructive prediction methodologies, namely ANN and SVM to determine number and weight measurements in pomegranate arils based on physical attributes.
ANN had the highest rate of accuracy for aril weight among models, and both ANN and SVM were able to work well with nonlinear relationships hence producing reliable predictions.
Results of PCA analysis also revealed that fruit weight, stem end diameter and crown diameter were the most important predictors of the model.
The results suggest that machine learning has the power to enhance pomegranate quality assessment, grading and postharvest management thus supporting its application in precision horticulture.
In this study, ‘Wonderful’ pomegranate fruits were used as materials and the prediction of the outputs (number of arils and arils weight) had been done with five inputs namely fruit weight (g), fruit width (mm), fruit height (mm), fruit crown diameter (mm) and stem end diameter (mm). Predictions of the outputs were done with four different methods including multiple linear regression (MLR), principal component analyses (PCA), artificial neural network (ANN) and support vector machines (SVM). The findings of this study provide evidence for the ability of non-destructive prediction methodologies, namely ANN and SVM to determine number and weight measurements in pomegranate arils based on physical attributes.
ANN had the highest rate of accuracy for aril weight among models, and both ANN and SVM were able to work well with nonlinear relationships hence producing reliable predictions.
Results of PCA analysis also revealed that fruit weight, stem end diameter and crown diameter were the most important predictors of the model.
The results suggest that machine learning has the power to enhance pomegranate quality assessment, grading and postharvest management thus supporting its application in precision horticulture.
Authors
E. Çağlar, İ. Kahramanoğlu
Keywords
modeling, multiple linear regression (MLR), principal component analyses (PCA), artificial neural network (ANN), support vector machines (SVM), artificial intelligence technologies and management
Online Articles (49)