Smart Detection of Sugarcane Leaf Rust Using UAV-Based Multispectral Imaging and Mask R-CNN

Abstract

Detecting and monitoring sugarcane leaf rust disease remains a significant challenge in precision agriculture due to the complex symptoms and changing environmental conditions. This study introduces a smart detection framework that combines Unmanned Aerial Vehicles (UAVs) equipped with multispectral imaging with a deep learning-based Mask R-CNN model to identify and segment leaf rust-infected areas on sugarcane leaves. The system was trained and validated using annotated multispectral images captured from UAV footage at altitudes of 5, 6, and 7 meters above ground. Performance evaluation, based on precision, recall, and F1 score, revealed a clear inverse relationship between detection accuracy and flight altitude. The highest F1 Score of 0.70 was recorded at 5 meters, dropping to 0.48 at 6 meters and 0.30 at 7 meters. These results demonstrate that combining Mask R-CNN with UAV-based multispectral data has strong potential for precise and scalable plant disease detection. Additionally, the study highlights key factors that impact segmentation accuracy, including image resolution determined by UAV altitude, natural shadow interference, and overlapping leaf structures. These findings support the development of UAV- and deep learning-based plant disease detection systems, especially in tropical agricultural environments, by enabling efficient, accurate, and field-adaptive monitoring of sugarcane leaf rust disease.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint