Abstract:
Avocados have become a popular fruits among health- conscious consumers. People often choose ripe avocados based on the skin color of the avocado peel or from squeezing the avocado causing the flesh to bruise. Therefore, this research aimed to investigate the relationship of firmness with the color value of avocado peel, and to measure the ripeness of avocado fruits using image processing method and a mathematical model to predict the ripeness level of avocados.
Image processing techniques were applied in combination with the multi-layer perceptron (MLP), Support Vector Regression (SVR), and K- Nearest Neighbors (KNN) models to predict the ripeness of avocado fruits. The experiments started with extracting the color value from avocado peels using image processing in the L*a*b* color space. The extracted data were sent to a model that splits the training data and testing data into the ratios of 70/30, 80/20, and 90/10.
The results revealed that the most suitable model for predicting avocado ripeness was the MLP model (90/10) with R2, RMSE and MAE values of 0.990, 0.341, and 0. 321, respectively. The R2, RMSE and MAE values of KNN model (80/ 20) were 0. 787, 6.953, and 1.672, respectively. The R2, RMSE and MAE values of SVR model (90/10) were 0. 776, 8. 310, and 1. 993, respectively. The MLP model could accurately predict up to 99.4%, while the KNN and SVR models were 95.2 and 94.9%, respectively. These results can be useful in developing models for predicting the ripening of avocados.