<![CDATA[Coffee is one of Indonesia’s major agricultural commodities, yet its productivity is often limited by inefficient fertilizer management, particularly in determining nitrogen (N), phosphorus (P), and potassium (K) requirements. Although conventional soil and leaf analyses are reliable, they are time-consuming and less practical for smallholder farmers. This underscores the need for an accurate, scalable, and cost-effective solution to optimize fertilizer usage. To address this issue, the study introduces an AI-based predictor for assessing NPK sufficiency in coffee plants. The research integrates computer vision and metaheuristic optimization to form a practical decision-support system. A dataset containing 12,000 images of coffee leaves was classified into three categories: Deficient, Sufficient, and Excessive. Image preprocessing involved resizing, grayscale conversion, HSV transformation, and normalization. Feature extraction utilized Histogram of Oriented Gradients (HOG) and HSV Color Histograms, followed by classification using a Support Vector Machine (SVM) optimized with the Whale Optimization Algorithm (WOA). The model achieved an accuracy exceeding 97%, effectively recognizing Deficient and Sufficient categories, with most misclassifications occurring in the Excessive class due to visual similarities. Model performance was validated using a confusion matrix, learning curve, and PCA visualization, confirming efficient convergence. The study highlights the promise of AI-driven solutions in enhancing precision agriculture and promoting sustainable coffee farming practices.]]>
Copyrights © 2025
