Developing prediction model for ripeness level of melon (Cucumis melo L.) based on acoustic impulses using artificial neural network (ANN)

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Non-climacteric fruits such as melons are often difficult to assess for ripeness due to the lack of obvious physical changes upon ripeness. Currently, melon producers rely on visual assessment, which requires specialized expertise and involves inefficient, destructive test methods. Thus, a non-destructive approach like acoustic impulse is crucial for ripeness classification. This study aims to develop a melon ripeness prediction model based on an acoustic impulse test. A total of 120 Honey Globe melons (Cucumis melo var. inodorus) were measured for acoustic properties (dominant frequency, magnitude, zero moment power, and short-term energy), using a knocker to obtain acoustic parameter data. The acoustic parameter data were processed using an artificial neural network (ANN) with the Backpropagation method, and model performance was evaluated using a confusion matrix. The model showed excellent performance with an R2 value of 94.46% and accuracy, precision, and recall rates of 98.33%, 97.37%, and 97.37%, respectively.