Development of a Non-Destructive Fruit Quality Predictor Using Convolutional Neural Network Regression Model on Raspberry Pi

Abstract

The rapid progress in agriculture continually increases the demand for high-quality fruits. However, the quality of the fruits was usually assessed by a destructive method that normally led to a concern that this traditional approach is time-consuming and prone to human mistakes. To address these problems, an automatic non-destructive quality assessment of fruits using a Convolutional Neural Network (CNN) regression model based on images will be developed. Ficus Carica L. (figs) was used as a sample validation for the quality predictor. Three different stages (stage 1, stage 2, and stage 3) of fig fruit images were captured using a Pi camera and implemented in raspberry Pi3 model B+. The developed model will be used to predict the internal quality of fig fruits. In this paper, the quality of fig fruit was assessed in terms of its chemical properties which is the Brix sugar value (%). The model has been trained and validated to have a Root Mean Square Error (RMSE) of 1.38. Hence, the quality predictor is based on this prediction model and the performance was tested to give an RMSE of 0.99.