Smart Harvesting with AI: Analyzing Orange Ripeness Using Convolutional Neural Networks and Support Vector Machines

Detection of orange ripeness is an important part of modern precision agriculture to achieve optimum harvesting, decreased postharvest losses and better orange quality. In this paper, based on the feature extraction ability of Convolutional Neural Networks (CNNs) and superior classification power of Support Vector Machines (SVMs), we present a hybrid Convolutional Neural Network - Support Vector Machine (CNN - SVM) framework for automated orange ripeness classification. The proposed model was trained and tested on a custom dataset of 5,000 labeled orange images, categorized into three ripeness stages: unripe, ripe, and overripe. Experimental results demonstrated that the CNN-SVM framework achieved an impressive classification accuracy of 97.5%, significantly outperforming standalone CNN (accuracy: 95. Our proposed approach achieved an accuracy of 85.6% and outperformed baseline eyewitness based models (accuracy: 76.2%), Support Vector Machine Convolutional Neural Networks (accuracy: 81.8%) and traditional SVM models (accuracy: 85.4%). The CNN-SVM model achieved precision and recall values of 96.7% and 97.3%, respectively, for distinguishing ripeness categories with minimum misclassifications, and comparison of the model with alternative methods like LSTM, Federated Learning CNN, CNN-RF and, RF validated its effectiveness in classifying ripeness categories. Our proposed approach also displayed a high area under the curve (AUC) value of 0.98, indicating potent discriminatory power. The analysis of confusion matrix shows very minimal misclassification and occurs between adjacent ripeness stages. Findings show the CNN-SVM framework represents an efficient, scalable, and accurate method for real time ripeness detection and actionable insights for the farmer to optimize harvesting decisions. As the future research, the model will be deployed to real world scenarios through IoT based systems in order to improve smart agricultural practices and increase productivity in citrus industry.