Classification of cherry maturity stages using machine learning methods

Accurate and rapid determination of the maturity level of agricultural products is of great importance for identifying the appropriate harvest time, preserving product quality, and minimizing economic losses. In particular, in fruits with high export potential such as cherries, determining the maturity stage accurately is a critical process for both producers and consumers. In this study, a machine learning-based method is proposed for classifying different maturity stages of cherries. Within the scope of the study, cherries belonging to five different maturity stages were collected from cherry orchards in Elazığ province, and a total of 3000 high-resolution images were obtained. The images were subjected to preprocessing steps, followed by feature extraction, and a dataset was constructed for classification. Ten different machine learning algorithms were employed in the experiments. These algorithms include Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), XGBoost, Logistic Regression, Ridge Classifier, LightGBM, CatBoost, Random Forest, Extra Trees, and K-Nearest Neighbors (KNN). The models were evaluated both without optimization and with an Optuna-based optimization process. Experimental findings demonstrated that optimization significantly improved classification performance. Among the models, the SVM achieved the highest accuracy rate compared to the others. The accuracy value obtained without optimization was 93.33%, while this value increased to 95.16% after optimization. In addition, other machine learning methods also achieved high accuracy rates, and in particular, some models showed a significant improvement in classification success after optimization. The results indicate that machine learning-based approaches are highly effective in classifying cherry maturity stages. These methods are considered to contribute to the improvement of quality control, harvest planning, and the development of intelligent decision support systems in agricultural production. Furthermore, the findings of this study are expected to shed light on future research to be conducted on different fruit types and to support digitalization processes in agricultural production.