Classification of pH Status in Stevia Plants Using Supervised Neural Networks and Computer Vision

Abstract

Artificial Intelligence and Computer Vision have revolutionized Precision Agriculture, enabling automated crop monitoring. This study proposes a model based on Neural Networks to classify soil pH in Stevia rebaudiana Bertoni crops, optimizing agricultural management and crop sustainability. To achieve this, Stevia images were processed using data augmentation techniques, extracting color features in RGB and hexadecimal formats. A supervised Artificial Neural Network was then trained to classify soil pH into acidic, optimal, and alkaline categories. The proposed model, StePHVIA, achieved 99% accuracy, outperforming pretrained architectures such as MobileNetV2 (97.38%) and ResNet-50 (76.38%). The evaluation was conducted using metrics such as Matthews Correlation Coefficient, accuracy, recall, and F1-score. These results confirm the effectiveness of Computer Vision and Deep Learning in Precision Agriculture, providing a real time and low cost alternative for soil monitoring. StePHVIA facilitates the early detection of soil imbalances, optimizing fertilizer application and improving Stevia crop productivity.