A Non-Destructive Approach of Young Coconut Maturity Detection using Acoustic Vibration and Neural Network

Fruit maturity classification particularly coconut could bring significant impact in improving the post-harvest industry standards. One of the major reason is because the maturity is related with quality and could affect the commercialization of the product. In the Philippines setting, young coconut or commonly known as buko are classified into three stages namely: Malauhog, Malakanin and Malakatad. In this paper, a classification system is proposed in order to establish a more scientific method of predicting the maturity stage of young coconut. Sound vibrations were created by a constant source of vibration from a vibration motor then the sound signatures or acoustic response were collected by the vibration sensor. The vibrations were collected three (3) times from the three (3) faces and three (3) edges of each samples, hence resulting to eighteen (18) sound signatures for every coconut. The model was trained and tested using 150 samples. Nine (9) sound signatures out of the eighteen was identified to be significant in the training phase after PCA has been applied. There were four (4) classifiers used in the study namely: SVM, KNN, ANN and DT. The classification system is based on the prediction model with the highest accuracy which is the Artificial Neural Network (ANN). The ANN model obtained an accuracy 93.3%.