Predicting maturity of coconut fruit from acoustic signal with applications of deep learning

Currently, a lot of attention is being paid to the evaluation and classification of horticulture crops, especially fruits. Maturity prediction is a major step in deciding the value of a coconut (Cocos nucifera), which is directly linked to the quality of the product. The sound-based deep learning method used to predict the maturity of a coconut can also be greatly beneficial to a number of tropical countries like the Philippines, Indonesia, and India who produce and export coconuts worldwide due to its high demand. A way to accurately determine the maturity level of a coconut is essential, as it affects the benefits that the fruit will provide.

This paper aims to develop an effective AI-driven method to predict the maturity level of a coconut using acoustic signals. The proposed sound-based autonomous approach exploits various deep learning models including customized CNN, pre-trained networks, i.e. the ResNet-50, VGG-16, VGG-19 and Inception V3 models for maturity level classification of the coconuts. The proposed study also demonstrates the usefulness of various deep learning models in inspecting coconuts and providing a promising accuracy level to automatically predict the maturity of coconuts into three classes, i.e. pre-mature, mature, and overripe coconuts, by using a small amount of input acoustic data. We have used an open access dataset that consists of a total of 381 raw acoustic signals, which is the result of knocking 127 coconut samples on its three ridges namely ``Ridge A’’, ``Ridge B’’, and ``Ridge C’’. Promising results are obtained by the proposed method of coconut maturity prediction, enabling producers to accurately determine the yield and quality of the product.