Deep Learning based Non-Destructive Evaluation of Physiochemical Parameters for Post Harvested Pomegranate Fruits

Abstract

In recent years, there has been an increasing consumer demand for fresh fruits of improved quality. Nowadays, pomegranate fruits attained increased popularity because of their high nutritional values and pharmacological characteristics. In order to assess the fruit's quality, the prediction of physiochemical parameters is necessary. Thus, the proposed work utilized a deep learning model for prediction analysis. The proposed work has studied a non-destructive determination of physiochemical parameters for pomegranate fruits. The internal images of the pomegranate fruit are obtained using magnetic resonance imaging (MRT). Then, the fruit quality is determined by predicting the physiochemical parameters like TSS, pH, acidity and firmness by combining the features of physiochemical and the GLCM (Gray Level Co-occurrence Matrix). The proposed study used the LSTM (Long-Short Term Memory) model and Stacked Dense Deep LSTM (SDD-LSTM) for prediction purposes. The performance of both the LSTM and SDD-LSTM based prediction models is measured by evaluating the metrics like a square of the correlation coefficient (r 2 ), root mean square error of calibration (RMSEC), and root mean square error of prediction (RMSEP). Also, the obtained results of the proposed prediction models, along with the combined features, are compared with other machine learning techniques. The result analysis shows that the proposed SDD-LSTM and GLCM have a higher potential for non-destructive assessment of physicochemical values of pomegranate fruit, which may be helpful in fruit post-harvest management.