Abstract
Iron levels play a crucial role in fruit quality and long-term productivity in peach orchards. Traditional diagnostic techniques, while effective, are often costly, time-consuming, and reliant on expert intervention. In response to these limitations, this study introduces a novel and cost-efficient methodology for assessing iron status in peach trees in the Pinggu region of China. The proposed approach combines digital imaging and artificial neural networks (ANN) to improve the accuracy of leaf iron evaluation. A total of 832 leaf samples were collected, and their active iron content (Fe²⁺) was measured using standard laboratory procedures. High-resolution leaf images were captured and analyzed across RGB, HSV, and CIE Lab color spaces. Feature extraction and dimensionality reduction were performed using Principal Component Analysis (PCA), and classification of iron levels was conducted using a k-nearest neighbors (KNN) algorithm within a PCA-optimized ANN framework. The resulting model, leveraging six principal components, achieved an overall classification accuracy of 86.7%, with precision, recall, and F1-scores exceeding 86.5% for all classes. Receiver Operating Characteristic (ROC) curve analysis further confirmed the robustness of the model. Moderate iron levels were more prone to misclassification, typically confused with either low or severe levels, highlighting the challenge of moderate differentiation. These results suggest that the proposed framework has considerable potential for evaluating iron status in peach leaves and may serve as a cost-effective and scalable approach for supporting precision agriculture practices.