Phenological States Classification of Blueberries Using Artificial Intelligence

The crop of blueberries in Colombia has experienced sustained growth in recent years, favoured by its geographical location, which allows fruit production throughout the year. A critical aspect of production is the classification of the plant's reproductive phenological stages, as this information is essential for accurately predicting the harvesting date of the ripe fruit. Currently, this process is carried out empirically, based on farmers experience, which can lead to errors in farm planning. For this reason, this study proposes an approach based on artificial intelligence algorithms, specifically convolutional neural networks (CNN), to assist farmers and agronomists in the automated classification of the reproductive phenological stages of the Biloxi variety. The model was trained on a dataset of approximately 8,000 RGB images representing nine different phenological stages of the plant. As a result, the model achieved 99% classification accuracy, representing a significant advance in automating phenological monitoring and optimising crop management. Once the artificial intelligence model was developed, it was implemented in the MaixCam device in order to classify the states in real time.