Precision agriculture in oil palm: A PCA and SCR-1DResNet-based classification approach

Determining plant health, particularly in oil palm (Elaeis guineensis), is essential for sustaining agricultural productivity, preventing yield loss, and enabling timely intervention against diseases and nutrient deficiencies. Conventional manual diagnosis is often time-consuming, subjective, and impractical for large-scale plantations. To address these limitations, this study proposes a hybrid classification framework that integrates Principal Component Analysis (PCA) for dimensionality reduction with a Self-Calibrated Residual 1D Convolutional Network (SCR-1DResNet) combined through Random Forest–based decision fusion. The proposed model is evaluated on a multiclass oil palm image dataset comprising five categories: Brown Spot, Healthy, White Scale, Nitrogen Deficiency, and Potassium Deficiency. Experimental results demonstrate that the proposed approach consistently outperforms several state-of-the-art deep learning models, including CNN-AutoMic, CNN-EML, Gabor-CNN, CNN-AdaBoost, and EB-CNN. On the testing dataset, the proposed method achieves an average accuracy exceeding 94%, with corresponding improvements in precision (94%), F1-score (87.5%), recall (83.5%), and specificity (above 83%) across all classes. These performance gains highlight the effectiveness of combining PCA-based feature compression with self-calibrated residual learning, particularly under class-imbalanced conditions. The results confirm the robustness, scalability, and practical applicability of the proposed framework for reliable plant health assessment, contributing to the advancement of precision agriculture in oil palm cultivation. • A hybrid image analysis method improves oil palm leaf health classification • Feature reduction enhances efficiency while preserving important visual patterns • Combined learning models increase robustness under imbalanced data conditions • The approach achieves high accuracy across diseases and nutrient deficiencies • The method supports reliable and scalable precision agriculture applications