<![CDATA[Leaf blight disease remains one of the major threats to rice production in Indonesia, causing significant yield losses and threatening national food security. Conventional detection methods rely heavily on manual field inspection, which is time-consuming, labor-intensive, and often ineffective for early-stage identification. Recent advances in multispectral drone imagery and artificial intelligence (AI) offer new opportunities for precision agriculture by enabling rapid, accurate, and large-scale crop health monitoring. However, the practical application of these technologies in Indonesian rice paddies is still limited and requires empirical validation. This study aims to examine the effectiveness of multispectral drone imagery integrated with AI-based classification models for the early detection of leaf blight disease in Indonesian rice fields. The research focuses on improving detection accuracy and supporting timely disease management decisions for farmers and agricultural stakeholders. The study employs an experimental research design using multispectral drone data collected from rice paddies in West Java during the growing season. Vegetation indices such as NDVI and GNDVI were extracted and analyzed using machine learning algorithms, including Random Forest and Convolutional Neural Networks (CNN). Ground truth data were obtained through field observations and laboratory confirmation to validate the model outputs. The results demonstrate that the AI-based model achieved high classification accuracy, exceeding 90% in detecting early-stage leaf blight symptoms. The integration of multispectral data significantly improved detection performance compared to visual RGB imagery alone. The study concludes that multispectral drone imagery combined with AI provides a reliable and efficient approach for early detection of leaf blight disease in rice paddies. This approach has strong potential to support precision agriculture, reduce crop losses, and enhance sustainable rice production in Indonesia.]]>
Copyrights © 2025
