<![CDATA[Oil palm disease detection remains a significant challenge in precision agriculture due to variability in dataset quality, image acquisition conditions, and deep learning model performance. This study presents a systematic review and meta-analysis of deep learning architectures, preprocessing techniques, and dataset characteristics used in oil palm disease classification. The review followed the PRISMA framework, identifying 165 articles from IEEE Xplore, SpringerLink, and ScienceDirect. After applying eligibility and quality criteria, 11 studies were included for qualitative and quantitative synthesis. The analyzed methods include Convolutional Neural Networks (CNN), ResNet, EfficientNet, U-Net, Artificial Neural Networks (ANN), and hybrid models integrating preprocessing techniques such as CLAHE and Median Blur. Reported accuracy ranged from 83.21% to 99.67%. Meta-analytical findings indicate that dataset diversity, class balance, and image enhancement significantly influence performance, with preprocessing contributing to accuracy improvements of up to approximately 5%. This study provides the first focused synthesis on oil palm disease classification. It highlights the importance of standardized datasets and robust methodological design to enhance reproducibility and support precision agriculture informatics.]]>
Copyrights © 2026
